Processing, storing, and transmitting information accounts for ∼10% of global energy use; projections suggest that computational energy demands will be 10× higher than the projected global energy supply by 2040. Realizing solid-state analogs of neural circuitry, using ‘neuromorphic’ materials, holds promise for enabling a new energy-efficient computing paradigm. The metal–insulator transitions (MITs) of electron-correlated transition-metal oxides provide an attractive vector for achieving large conductance switching with minimal energy dissipation. Here, we review current understanding of the mechanisms underpinning electronic instabilities, discuss methods for modulation of spiking behavior through tuning of atomistic and electronic structure, and highlight the need for establishing deterministic and independent control of transformation characteristics such as switching magnitude, energy thresholds, heat dissipation, hysteresis, and dynamics of relaxation.

处理，存储和传输信息约占全球能源使用量的10％；预测表明，到2040年，计算能源需求将比预计的全球能源供应高10倍。使用“神经形态”材料实现神经电路的固态模拟，有望实现新的节能计算范式。电子相关的过渡金属氧化物的金属-绝缘体转变（MIT）提供了一个诱人的载体，可实现以最小的能量消耗实现大的电导率转换。在这里，我们回顾了目前对构成电子不稳定性的机制的理解，讨论了通过调整原子和电子结构来调节尖峰行为的方法，