From a Single Dendrite to Dendritic Networks: Towards Multi-Terminal OECTs

In the quest to change the way we envision computing, the ability to create connections in between computing nodes on demand offers the very exciting possibility to explore bottom-up strategies when it comes to thinking the design of electronic devices and circuits.<br/>Recently, the growth of conductive polymer fibers via electropolymerization (most notably poly(3,4-ethylenedioxythiophene) doped with poly(styrenesulfonate), abbreviated as PEDOT:PSS) was employed in the realm of neuromorphic hardware, first as a way to tune the resistance of a connection,[1] and then to perform computing tasks, such as biosignal classification trough reservoir computing,[2] thus establishing the advantages of the method.<br/><br/>In this work, we propose to demonstrate that electropolymerization is a useful tool for the creation of sensors that are able to realize computing tasks in an aqueous environment. By taking advantage of the morphology of the fibers, we show that PEDOT:PSS dendrites can discriminate between different types of voltage pulses emitted in their vicinity by a local gate electrode, thus performing <i>in materio</i> classification.[3]<br/>In addition, we discuss the growth of networks of polymer fibers on 2D substrates. The ability to create structures that cover several microelectrodes allows us to study the behavior of the whole system instead of a single dendrite, and it highlights the relationship that relates the morphology of the object and its electrical properties. In particular, we show that dendrites present an asymmetric non-linear behavior due to the volume of polymer that increases the capacitance of the device albeit not participating in conduction. Moreover, because of the ionoelectronic coupling that exists within an electrolyte, two active devices working concomitantly will influence one another. We demonstrate that it is a blessing for the realization of computing tasks, such as logic, and that it can also be used to program the dendrites into non-volatile conductance states. Therefore, dendritic networks could constitute a new building block for non-conventional information processing that fits into the larger framework of brain-inspired computing.<br/><br/><b>References</b><br/>[1] M. Akai-Kasaya <i>et al.</i>, “Evolving conductive polymer neural networks on wetware,” <i>Jpn. J. Appl. Phys.</i>, vol. 59, no. 6, p. 060601, Jun. 2020, doi: 10.35848/1347-4065/ab8e06.<br/>[2] M. Cucchi <i>et al.</i>, “Reservoir computing with biocompatible organic electrochemical networks for brain-inspired biosignal classification,” <i>Sci. Adv.</i>, vol. 7, no. 34, p. eabh0693, Aug. 2021, doi: 10.1126/sciadv.abh0693.<br/>[3] C. Scholaert, K. Janzakova, Y. Coffinier, F. Alibart, and S. Pecqueur, “Plasticity of conducting polymer dendrites to bursts of voltage spikes in phosphate buffered saline,” <i>Neuromorphic Comput. Eng.</i>, vol. 2, no. 4, p. 044010, Dec. 2022, doi: 10.1088/2634-4386/ac9b85.