MXene‑Ti3C2Tx‑based neuromorphic computing: physical mechanisms, performance enhancement, and cutting‑edge computing

A groundbreaking review published in Nano-Micro Letters provides a comprehensive overview of MXene-Ti₃C₂T_x as a universal platform for neuromorphic devices. Authored by Yubo Fan, the review highlights how this single 2-D material overcomes the limitations of traditional CMOS and sustains the scaling trajectory in the post-Moore era.

Why This Research Matters

• Overcoming CMOS Bottlenecks: As AI workloads surge, conventional architectures suffer from >100 pJ per MAC and millisecond latency due to sensor-memory-processor separation. MXene-Ti₃C₂T_x monolayers integrate all three functions, delivering sub-femtojoule synaptic events and nanosecond response at sub-1 V operation.  

• Enabling More-than-Moore Applications: From 5G/6G edge intelligence to implantable neuro-prosthetics, MXene devices enable flexible, multimodal, and biocompatible systems that classical silicon cannot match.

Innovative Design and Mechanisms

• MXene-Ti₃C₂T_x for Neuromorphic Devices: The review systematically covers four core switching mechanisms—ECM, VCM, electron tunneling, and charge trapping—each mapped to specific device stacks (Ag/Ti₃C₂T_x/Pt, TiOx/Ti₃C₂T_x/Au, etc.).  

• Advanced Device Structures: Steep-slope TFETs, NCFETs, and floating-gate transistors built from MXene exhibit on/off ratios >10⁶, sub-0.5 V SET, and endurance >10⁴ cycles—outperforming incumbent RRAM and FeFET technologies.  

• 3D Integration & Flexible Electronics: Solution-processable MXene inks enable wafer-level spin-coating, 3-D monolithic stacking, and roll-to-roll fabrication on polyimide, glass, or biodegradable substrates for bendable e-skins and retinal implants.

Applications and Future Outlook

• In-Memory Computing and Neuromorphic Arrays: Crossbar circuits demonstrate 96.4 % MNIST accuracy at 0.35 pJ per inference, while 1S-1N networks encode grayscale images into spike-timing with 20× fewer training epochs than GPU baselines.  

• Multimodal Sense-Compute Nodes: Tactile, optical, and neurotransmitter sensors directly modulate synaptic weights—achieving 80 % material recognition via glove-mounted arrays and 1 aM-level dopamine detection for closed-loop therapeutics.  

• Future Research Directions: Priorities include wafer-scale defect control, CMOS-BEOL-compatible patterning, and biodegradable encapsulation to accelerate clinical translation.

Conclusions
This review provides a complete blueprint—from atomic mechanisms to system demonstrations—showing how MXene-Ti₃C₂T_x redefines neuromorphic hardware. By collapsing sensing, memory, and computation into one atomic layer, the material promises ultra-low-power, highly integrated circuits that push beyond Moore’s Law while merging seamlessly with biological systems.