In recent years the field of neuromorphic low-power systems that consume orders of magnitude less power gained significant momentum. However, their wider use is still hindered by the lack of algorithms that can harness the strengths of such architectures. While neuromorphic adaptations of representation learning algorithms are now emerging, efficient processing of temporal sequences or variable length-inputs remain difficult. Recurrent neural networks (RNN) are widely used in machine learning to solve a variety of sequence learning tasks. In this work we present a train-and-constrain methodology that enables the mapping of machine learned (Elman) RNNs on a substrate of spiking neurons, while being compatible with the capabilities of current and near-future neuromorphic systems. This "train-and-constrain" method consists of first training RNNs using backpropagation through time, then discretizing the weights and finally converting them to spiking RNNs by matching the responses of artificial neurons with those of the spiking neurons. We demonstrate our approach by mapping a natural language processing task (question classification), where we demonstrate the entire mapping process of the recurrent layer of the network on IBM's Neurosynaptic System "TrueNorth", a spike-based digital neuromorphic hardware architecture. TrueNorth imposes specific constraints on connectivity, neural and synaptic parameters. To satisfy these constraints, it was necessary to discretize the synaptic weights and neural activities to 16 levels, and to limit fan-in to 64 inputs. We find that short synaptic delays are sufficient to implement the dynamical (temporal) aspect of the RNN in the question classification task. The hardware-constrained model achieved 74% accuracy in question classification while using less than 0.025% of the cores on one TrueNorth chip, resulting in an estimated power consumption of ~17 uW.

We present an approach to constructing a neuromorphic device that responds to language input by producing neuron spikes in proportion to the strength of the appropriate positive or negative emotional response. Specifically, we perform a fine-grained sentiment analysis task with implementations on two different systems: one using conventional spiking neural network (SNN) simulators and the other one using IBM's Neurosynaptic System TrueNorth. Input words are projected into a high-dimensional semantic space and processed through a fully-connected neural network (FCNN) containing rectified linear units trained via backpropagation. After training, this FCNN is converted to a SNN by substituting the ReLUs with integrate-and-fire neurons. We show that there is practically no performance loss due to conversion to a spiking network on a sentiment analysis test set, i.e. correlations between predictions and human annotations differ by less than 0.02 comparing the original DNN and its spiking equivalent. Additionally, we show that the SNN generated with this technique can be mapped to existing neuromorphic hardware -- in our case, the TrueNorth chip. Mapping to the chip involves 4-bit synaptic weight discretization and adjustment of the neuron thresholds. The resulting end-to-end system can take a user input, i.e. a word in a vocabulary of over 300,000 words, and estimate its sentiment on TrueNorth with a power consumption of approximately 50 uW.

Increased interactions at the graphene-metal interface are here demonstrated to yield an effective prevention of intercalation of foreign species below the graphene cover. Hereby, an engineering pathway for increasing the usability of graphene as a metal coating is demonstrated. Graphene on Ir(111) (Gr/Ir(111)) is used as a model system, as it has previously been well-established that an increased interaction and formation of chemical bonds at the graphene-Ir interface can be induced by hydrogen functionalization of the graphene from its top side. With X-ray photoelectron spectroscopy, it is shown that hydrogen-induced increased interactions at the Gr/Ir(111) interface effectively prevents intercalation of CO in the millibar range. The scheme leads to protection against at least 10 times higher pressure and 70 times higher fluences of CO, compared to the protection offered by pristine Gr/Ir(111).

Hydrogen functionalization of graphene by exposure to vibrationally excited H₂ molecules is investigated by combined scanning tunneling microscopy, high-resolution electron energy loss spectroscopy, X-ray photoelectron spectroscopy measurements, and density functional theory calculations. The measurements reveal that vibrationally excited H₂ molecules dissociatively adsorb on graphene on Ir(111) resulting in nanopatterned hydrogen functionalization structures. Calculations demonstrate that the presence of the Ir surface below the graphene lowers the H₂ dissociative adsorption barrier and allows for the adsorption reaction at energies well below the dissociation threshold of the H-H bond. The first reacting H₂ molecule must contain considerable vibrational energy to overcome the dissociative adsorption barrier. However, this initial adsorption further activates the surface resulting in reduced barriers for dissociative adsorption of subsequent H₂ molecules. This enables functionalization by H₂ molecules with lower vibrational energy, yielding an avalanche effect for the hydrogenation reaction. These results provide an example of a catalytically active graphene-coated surface and additionally set the stage for a re-interpretation of previous experimental work involving elevated H₂ background gas pressures in the presence of hot filaments.