Fig. 1: Design of the mode-switchable cv-OECT.

a, Comparison between the biological nervous system and cv-OECT-based artificial nervous system, where cv-OECT can act as both volatile receptor and non-volatile synapse. Optical micrographs display the top view of a v-OECT (scale bar, 100 μm). b, Device architecture of v-OECT; the two dashed boxes show the ion contribution in the volatile/non-volatile mode and the chemical structure of PTBT-p, respectively. c, Cryo-EM images of the 200 °C-thermal annealed (TA) and as-cast PTBT-p films. d, Transfer curves of cv-OECT with polarizable/non-polarizable gate electrode. e, Normalized 0–1 absorbance as a function of doping potential; the inset shows the setup for UV–vis measurement. Stages I and II correspond to the doping of amorphous and crystalline regions, respectively. f, Time-resolved UV–vis spectra of channels correspond well with the device performance. g, XPS spectra of as-cast and annealed p-OECT channels doped at LGP and HGP. The pink and blue lines are the signals from [TFSI⁻] before and after 30 nm etching. h, One-dimensional GIWAXS profile of the annealed film samples. Before measurement, the samples were doped at LGP or HGP and then grounded (Methods). Reversible displacement of the (100) peak between the high/low resistance state (HRS/LRS) suggests that the anions firmly embed among the glycol side chains in the crystalline region. i, Schematic explaining the mode-switching mechanism. The special channel dimensions and crystallization provide a high-barrier eV_b between the two ionic states (1 and 2), resulting in a non-volatile behaviour. V_b denotes the voltage bias that drives the ions to overcome the barrier. LGP can only inject ions into the amorphous regions and lead to volatile behaviour. When the non-polarizable gate was used, the counterions cannot be reduced on the gate and thus they migrate into and neutralize the channel because of the reversed electric field, making the device volatile.