Ionic floating-gate memories: Digital implementations of artificial neural networks perform many tasks, such as image recognition and language processing, but are too energy intensive for many applications. Analog circuits that use large crossbar arrays of synaptic memory elements represent a low-power alternative, but most devices cannot update the synaptic weights uniformly or scale to large array sizes. Fuller et al. developed an integrated device, ionic floating-gate memory, that has the gate terminal of a redox transistor electrically connected to a diffusive memristor. This low-power device enabled linear and symmetric weight updates in parallel over an entire crossbar array at megahertz rates over 10 ⁹ write-read cycles. Science , this issue p. 570

Devices with tunable resistance are highly sought after for neuromorphic computing. Conventional resistive memories, however, suffer from nonlinear and asymmetric resistance tuning and excessive write noise, degrading artificial neural network (ANN) accelerator performance. Emerging electrochemical random-access memories (ECRAMs) display write linearity, which enables substantially faster ANN training by array programing in parallel. However, state-of-the-art ECRAMs have not yet demonstrated stable and efficient operation at temperatures required for packaged electronic devices (~90°C). Here, we show that (semi)conducting polymers combined with ion gel electrolyte films enable solid-state ECRAMs with stable and nearly temperature-independent operation up to 90°C. These ECRAMs show linear resistance tuning over a >2× dynamic range, 20-nanosecond switching, submicrosecond write-read cycling, low noise, and low-voltage (±1 volt) and low-energy (~80 femtojoules per write) operation combined with excellent endurance (>10⁹ write-read operations at 90°C). Demonstration of these high-performance ECRAMs is a fundamental step toward their implementation in hardware ANNs.

A robot learns to follow a path to exit a maze through sensorimotor learning that is induced by an organic neuromorphic circuit.

In photovoltaic devices, the photo-generated charge carriers are typically assumed to be in thermal equilibrium with the lattice. In conventional materials, this assumption is experimentally justified as carrier thermalization completes before any significant carrier transport has occurred. Here, we demonstrate by unifying time-resolved optical and electrical experiments and Monte Carlo simulations over an exceptionally wide dynamic range that in the case of organic photovoltaic devices, this assumption is invalid. As the photo-generated carriers are transported to the electrodes, a substantial amount of their energy is lost by continuous thermalization in the disorder broadened density of states. Since thermalization occurs downward in energy, carrier motion is boosted by this process, leading to a time-dependent carrier mobility as confirmed by direct experiments. We identify the time and distance scales relevant for carrier extraction and show that the photo-generated carriers are extracted from the operating device before reaching thermal equilibrium.