We show the design and simulation of organic neuromorphic circuits in a hybrid-computation approach that emulates Boolean and reversible logic gates based on multigate organic electrochemical transistors (OECTs). The organic neuromorphic circuits consist of input, hidden, and output layers that can carry out Boolean operations, including the Exclusive OR (XOR) function, with five or less OECTs. The multigate functionality of OECTs is harnessed to perform the summation function of the neurons. Connection weights of the networks are defined in an unconventional way that depends on the value of the drain-source current of the outputting neuron, which changes according to the input values of the circuit. The Boolean circuits can be cascaded together to build higher level circuits and are demonstrated to form a full adder circuit and the Double Feynman and Toffoli reversible logic gates. Using realistic experimental parameters, the energy per computation is estimated to be ~2.3 nJ for circuit designs with a bias voltage of 0.5 V, with ~230 fJ or less being achievable for lower bias voltages.

我们在混合计算方法中展示了有机神经形态电路的设计和仿真，该方法模拟了基于多栅极有机电化学晶体管（OECT）的布尔和可逆逻辑门。有机神经形态电路由可以执行布尔运算的输入，隐藏和输出层组成，包括具有五个或更少的OECT的异或（XOR）功能。利用OECT的多门功能来执行神经元的求和功能。网络的连接权重以非常规方式定义，取决于输出神经元的漏源电流值，该值根据电路的输入值而变化。布尔电路可以级联在一起以构建更高级别的电路，并被演示形成完整的加法器电路以及Double Feynman和Toffoli可逆逻辑门。使用实际的实验参数，对于偏置电压为0.5 V的电路设计，每次计算的能量估计约为2.3 nJ，对于较低的偏置电压，则约为230 fJ或更低。