Chemical synapses are essential to the transmission of information through the central nervous system. When an electrical impulse – otherwise known as an action potential – reaches the end of a presynaptic axon, vesicles containing neurotransmitters bind to the plasma membrane and are released into the synaptic cleft. These neurotransmitters diffuse to the postsynaptic neuron where they lead to the generation of a postsynaptic current (PSC). The PSC does more than generate an action potential to shape the immediate actions of the postsynaptic neuron – it can also have a long-term effect on its cellular and molecular machinery.

To achieve this transducing behaviour in the FET, Das and his team equated the presynaptic action potential and the induced PSC with the gate voltage (VG) pulse and the change in source to drain current (ΔIDS), respectively. They also took advantage of the electron traps in the FET to display behaviour analogous to synaptic plasticity. This is the ability of synapses to strengthen and weaken dependent on the signalling activity, and forms one of the building blocks of memory formation in mammals.

Swapping vesicles for traps

To mimic neurotransmitter behaviour the researchers drew a parallel between vesicle exocytosis – the process of un-packaging the vesicles and releasing the neurotransmitters into the synapse – and the capture and release of electrons in the FET. They developed a system whereby the quantal, stochastic and excitatory/inhibitory releases of neurotransmitters were associated with the frequency, amplitude and polarity of VG pulses, respectively.

The electron traps in the FET cause a shift in the threshold voltage of the device such that when the gate voltage is pulsed negatively, ΔIDS is positive – a response akin to an excitatory neurotransmitter release. The exact mechanism of electron trapping is a subject of debate, with several origins posited by the authors:

• adsorption of gas molecules on the surface of MoS2;
• electron trapping at the interface between SiO2 and MoS2;
• and defects in the MoS2 itself – particularly the surface.

The FET also replicates PSC saturation. In a real synapse this saturation occurs due to the finite number of neurotransmitters that can be released into the synaptic cleft. In the FET – for a set gate voltage there are a fixed number of traps induced and as the number of VG pulses increases more traps become charged – at a higher number of pulses the threshold to charge all traps is reached and ΔIDS will saturate.

Synaptic plasticity

In excitatory synapses a long-lasting signal increase is called long-term potentiation (LTP) and is linked with learning. It can be replicated in the FET because of the comparatively large discharge time constant of the electron traps. By maximizing the pulse amplitude and the number of pulses as well as operating the device in the sub-threshold region – where any change in the threshold voltage will produce an exponential change in current – the researchers optimized the trap discharge time to achieve lengths similar to those of biological synapses.

The team fabricated the FET by depositing a mechanically exfoliated 2 nm-thick MoS2 flake – about three monolayers – on a Si/SiO2 substrate and e-beam evaporating Ni source–drain contacts. They claim the high-voltage operation of the FET – which raises issues such as high power consumption – can be easily resolved by substituting the SiO2 dielectric with ultra-thin or high-k dielectric materials.

Going forward the group plan to turn their attention to more complex issues in the field, including co-release of neurotransmitters as a mechanism for negative feedback to neural firing, and recurrent neural networks with slow nonlinear transmission.

Full details are reported in ACS Nano.