Figure 3.

Electrical modulation of ultralow power MoS₂‐based artificial synapse. a) Schematic of signal transmission from pre‐synapse to post‐synapse in biological synapse and emulation in MoS₂ device by electrical pulse. b) EPSC triggered by presynaptic pulse with different amplitudes (−3, −3.5, −4, −4.5, and −5 V). The pulse width was fixed at 100 ns. c) IPSC inhibited by different positive pulse (amplitudes: 3, 4, 5, 6, and 7 V) with the same pulse width of 100 ns. d) EPSC generated by a pair of pre‐spikes (−4 V, 100 ns) with an interval of 500 ms. e) PPF and PPD index (A₂/A₁ × 100%) as a function of interval time between two continuous pulses, fitted by exponential curves. f) Long‐term potentiation stimulated by a series of pre‐spikes (pulse amplitude = −4 V, pulse width = 100 ns, pulse interval = 10 s). g) Long‐term depression simulation: a series of pre‐spikes (5 V, 100 ns) at 10 s intervals. Inset presents the waveform used for simulation. h) Endurance of potentiation‐depression with 2000 consecutive spikes. i) Ultralow power during LTP and LTD simulation in MoS₂‐based synapse. ΔG _LTP is 1.146 nS with 100 negative pulses (−4 V, 100 ns) in LTP, where the average power is 18.3 aJ per spike. j) Long‐term retention behavior after applying 10 pulses (−4 V, 100 ns) for over 1000 s. k) Comparison of power consumption in biological and various electronic synapses. The level of attojoule power consumption is lower than ≈10 fJ of bio‐synapse.