Machine learning of atomic-scale properties is revolutionizing molecular modeling, making it possible to evaluate inter-atomic potentials with first-principles accuracy, at a fraction of the costs. The accuracy, speed, and reliability of machine learning potentials, however, depend strongly on the way atomic configurations are represented, i.e., the choice of descriptors used as input for the machine learning method. The raw Cartesian coordinates are typically transformed in "fingerprints," or "symmetry functions," that are designed to encode, in addition to the structure, important properties of the potential energy surface like its invariances with respect to rotation, translation, and permutation of like atoms. Here we discuss automatic protocols to select a number of fingerprints out of a large pool of candidates, based on the correlations that are intrinsic to the training data. This procedure can greatly simplify the construction of neural network potentials that strike the best balance between accuracy and computational efficiency and has the potential to accelerate by orders of magnitude the evaluation of Gaussian approximation potentials based on the smooth overlap of atomic positions kernel. We present applications to the construction of neural network potentials for water and for an Al-Mg-Si alloy and to the prediction of the formation energies of small organic molecules using Gaussian process regression.

At the interface with solids, the mobility of liquid molecules tends to be reduced compared with bulk, often resulting in increased local order due to interactions with the surface of the solid. At room temperature, liquids such as water and methanol can form solvation structures, but the molecules remain highly mobile, thus preventing the formation of long-lived supramolecular assemblies. Here we show that mixtures of water with methanol can form a novel type of interfaces with hydrophobic solids. Combining in situ atomic force microscopy and multiscale molecular dynamics simulations, we identify solid-like two-dimensional interfacial structures that nucleate thermally, and are held together by an extended network of hydrogen bonds. On graphite, nucleation occurs above ∼35 °C, resulting in robust, multilayered nanoscopic patterns. Our findings could have an impact on many fields where water-alcohol mixtures play an important role such as fuel cells, chemical synthesis, self-assembly, catalysis and surface treatments.

No abstract

The values of C 12 in Table I of the original article are wrong and the corrected table is shown below. The C 12 values shown in the original article are C 44 values and the C 12 values are corrected in this Table I. This correction does not affect the discussion and conclusion of the article.

We investigated a suspended bilayer graphene where the bottom/top layer is doped by boron/nitrogen substitutional atoms. By using density functional theory calculations, we found that at high dopant concentration (one B-N pair every 32 C atoms), the electronic structure of the bilayer does not depend on the B-N distance but on the relative occupation of the bilayer graphene sublattices by B and N. The presence of the dopants and the consequent charge transfer establish a built-in electric field between the layers, giving rise to an energy gap. We further investigated the electronic transport properties and found that intralayer current is weakly influenced by the presence of these dopants while the interlayer one is enhanced for biases, allowing an easy tunneling between layers. This effect leads to current rectification in asymmetric junctions. DOI: 10.1103/PhysRevB.93.205420 Bilayer graphene (BLG) is a two-dimensional material constituted by two stacked graphene layers. It has recently attracted much interest because it shows exceptionally high charge mobility like in single-layer graphene [1]. Charge transport in BLG is tied to the in-plane direction of each of the two layers and it has been shown that a gate potential can induce an energy gap in bilayer graphene [2]. As a consequence, BLG has been proposed to be more suitable than single-layer graphene to realize carbon-based field-effect logic devices, as the new bilayer-graphene-based transistor [3][4][5]. However, due to the small density of states at the Fermi level of BLG, very large electric fields are needed in order to create an unbalance in the charge distribution. This has a deleterious impact on the performance and stability of gated BLG. For instance, graphene and BLG deposited on oxide dielectrics such as SiO 2 show a large hysteresis in the I -V curve, due to the charging/discharging of point defects in the oxide. Moreover, charged impurities in the gate oxide and at the graphene-gate interface have been shown to dramatically reduce the mobility in graphene-based devices [6,7]. On the other hand, the low density of states can be exploited to induce the effects of a large electric field in BLG. In fact, the addition or removal of even a small amount of charge from each layer causes an abrupt change of the Fermi level.In this paper, we simulate a suspended bilayer graphene nanojunction where boron and nitrogen atoms substitute carbon atoms in the bottom and top layers, respectively. The equal number of B and N atoms assures that the system is isoelectronic to the pristine bilayer. By first-principles calculations, we find that a large built-in electric field is then established, due to the combination of two effects: the first is the Fermi-level mismatch between the isolated B-rich and N-rich layers, and the second is a partial charge transfer between them as they are put in contact. As a consequence, a small energy gap appears, enabling this system to be employed for nanoelectronics applications. In view of this, we also investigate the balli...