Biomolecular dynamics with machine-learned quantum-mechanical force fields trained on diverse chemical fragments.

The GEMS method enables molecular dynamics simulations of large heterogeneous systems at ab initio quality.

Molecules in Environments: Toward Systematic Quantum Embedding of Electrons and Drude Oscillators.

We develop a quantum embedding method that enables accurate and efficient treatment of interactions between molecules and an environment, while explicitly including many-body correlations. The molecule is composed of classical nuclei and quantum electrons, whereas the environment is modeled via charged quantum harmonic oscillators. We construct a general Hamiltonian and introduce a variational Ansatz for the correlated ground state of the fully interacting molecule-environment system. This wave function is optimized via the variational Monte Carlo method and the ground state energy is subsequently estimated through the diffusion Monte Carlo method. The proposed scheme allows an explicit many-body treatment of electrostatic, polarization, and dispersion interactions between the molecule and the environment. We study solvation energies and excitation energies of benzene derivatives, obtaining excellent agreement with explicit ab initio calculations and experiments.

"Freedom of design" in chemical compound space: towards rational in silico design of molecules with targeted quantum-mechanical properties.

The rational design of molecules with targeted quantum-mechanical (QM) properties requires an advanced understanding of the structure-property/property-property relationships (SPR/PPR) that exist across chemical compound space (CCS). In this work, we analyze these fundamental relationships in the sector of CCS spanned by small (primarily organic) molecules using the recently developed QM7-X dataset, a systematic, extensive, and tightly converged collection of 42 QM properties corresponding to ≈4.2M equilibrium and non-equilibrium molecular structures containing up to seven heavy/non-hydrogen atoms (including C, N, O, S, and Cl). By characterizing and enumerating progressively more complex manifolds of molecular property space-the corresponding high-dimensional space defined by the properties of each molecule in this sector of CCS-our analysis reveals that one has a substantial degree of flexibility or "freedom of design" when searching for a single molecule with a desired pair of properties or a set of distinct molecules sharing an array of properties. To explore how this intrinsic flexibility manifests in the molecular design process, we used multi-objective optimization to search for molecules with simultaneously large polarizabilities and HOMO-LUMO gaps; analysis of the resulting Pareto fronts identified non-trivial paths through CCS consisting of sequential structural and/or compositional changes that yield molecules with optimal combinations of these properties.

Equivariant Graph Neural Networks for Toxicity Prediction.

Predictive modeling of toxicity is a crucial step in the drug discovery pipeline. It can help filter out molecules with a high probability of failing in the early stages of de novo drug design. Thus, several machine learning (ML) models have been developed to predict the toxicity of molecules by combining classical ML techniques or deep neural networks with well-known molecular representations such as fingerprints or 2D graphs. But the more natural, accurate representation of molecules is expected to be defined in physical 3D space like in ab initio methods. Recent studies successfully used equivariant graph neural networks (EGNNs) for representation learning based on 3D structures to predict quantum-mechanical properties of molecules. Inspired by this, we investigated the performance of EGNNs to construct reliable ML models for toxicity prediction. We used the equivariant transformer (ET) model in TorchMD-NET for this. Eleven toxicity data sets taken from MoleculeNet, TDCommons, and ToxBenchmark have been considered to evaluate the capability of ET for toxicity prediction. Our results show that ET adequately learns 3D representations of molecules that can successfully correlate with toxicity activity, achieving good accuracies on most data sets comparable to state-of-the-art models. We also test a physicochemical property, namely, the total energy of a molecule, to inform the toxicity prediction with a physical prior. However, our work suggests that these two properties can not be related. We also provide an attention weight analysis for helping to understand the toxicity prediction in 3D space and thus increase the explainability of the ML model. In summary, our findings offer promising insights considering 3D geometry information via EGNNs and provide a straightforward way to integrate molecular conformers into ML-based pipelines for predicting and investigating toxicity prediction in physical space. We expect that in the future, especially for larger, more diverse data sets, EGNNs will be an essential tool in this domain.

Dithienylethene-Based Single Molecular Photothermal Linear Actuator.

By employing a mechanically controllable break junction technique, we have realized an ideal single molecular linear actuator based on dithienylethene (DTE) based molecular architecture, which undergoes reversible photothermal isomerization when subjected to UV irradiation under ambient conditions. As a result, open form (compressed, UV OFF) and closed form (elongated, UV ON) of dithienylethene-based molecular junctions are achieved. Interestingly, the mechanical actuation is achieved without changing the conductance of the molecular junction around the Fermi level over several cycles, which is an essential property required for an ideal single molecular actuator. Our study demonstrates a unique example of achieving a perfect balance between tunneling width and barrier height change upon photothermal isomerization, resulting in no change in conductance but a change in the molecular length, which results in mechanical actuation at the single molecular level.

QM7-X, a comprehensive dataset of quantum-mechanical properties spanning the chemical space of small organic molecules.

We introduce QM7-X, a comprehensive dataset of 42 physicochemical properties for ≈4.2 million equilibrium and non-equilibrium structures of small organic molecules with up to seven non-hydrogen (C, N, O, S, Cl) atoms. To span this fundamentally important region of chemical compound space (CCS), QM7-X includes an exhaustive sampling of (meta-)stable equilibrium structures-comprised of constitutional/structural isomers and stereoisomers, e.g., enantiomers and diastereomers (including cis-/trans- and conformational isomers)-as well as 100 non-equilibrium structural variations thereof to reach a total of ≈4.2 million molecular structures. Computed at the tightly converged quantum-mechanical PBE0+MBD level of theory, QM7-X contains global (molecular) and local (atom-in-a-molecule) properties ranging from ground state quantities (such as atomization energies and dipole moments) to response quantities (such as polarizability tensors and dispersion coefficients). By providing a systematic, extensive, and tightly-converged dataset of quantum-mechanically computed physicochemical properties, we expect that QM7-X will play a critical role in the development of next-generation machine-learning based models for exploring greater swaths of CCS and performing in silico design of molecules with targeted properties.

Accurate Many-Body Repulsive Potentials for Density-Functional Tight Binding from Deep Tensor Neural Networks.

We combine density-functional tight binding (DFTB) with deep tensor neural networks (DTNN) to maximize the strengths of both approaches in predicting structural, energetic, and vibrational molecular properties. The DTNN is used to construct a nonlinear model for the localized many-body interatomic repulsive energy, which so far has been treated in an atom-pairwise manner in DFTB. Substantially improving upon standard DFTB and DTNN, the resulting DFTB-NNrep model yields accurate predictions of atomization and isomerization energies, equilibrium geometries, vibrational frequencies, and dihedral rotation profiles for a large variety of organic molecules compared to the hybrid DFT-PBE0 functional. Our results highlight the potential of combining semiempirical electronic-structure methods with physically motivated machine learning approaches for predicting localized many-body interactions. We conclude by discussing future advancements of the DFTB-NNrep approach that could enable chemically accurate electronic-structure calculations for systems with tens of thousands of atoms.

Exploring the write-in process in molecular quantum cellular automata: a combined modelingand first-principle approach.

The molecular quantum cellular automata paradigm (m-QCA) offers a promising alternative framework to current CMOS implementations. A crucial aspect for implementing this technology concerns the construction of a device which effectively controls intramolecular charge-transfer processes. Tentative experimental implementations have been developed in which a voltage drop is created generating the forces that drive a molecule into a logic state. However, important factors such as the electric field profile, its possible time-dependency and the influence of temperature in the overall success of charge-transfer are relevant issues to be considered in the design of a reliable device. In this work, we theoretically study the role played by these processes in the overall intramolecular charge-transfer process. We have used a Landau-Zener (LZ) model, where different time-dependent electric field profiles have been simulated. The results have been further corroborated employing density functional tight-binding method. The role played by the nuclear motions in the electron-transfer process has been investigated beyond the Born-Oppenheimer approximation by computing the effect of the external electric field in the behavior of the potential energy surface. Hence, we demonstrate that the intramolecular charge-transfer process is a direct consequence of the coherent LZ nonadiabatic tunneling and the hybridization of the diabatic vibronic states which effectively reduces the trapping of the itinerant electron at the donor group.

Doping engineering of thermoelectric transport in BNC heteronanotubes.

BNC heteronanotubes are promising materials for the design of nanoscale thermoelectric devices. In particular, the structural BN doping pattern can be exploited to control the electrical and thermal transport properties of BNC nanostructures. We here address the thermoelectric transport properties of (6,6)-BNC heteronanotubes with helical and horizontal BN doping patterns. For this, we use a density functional tight-binding method combined with the Green's function technique. Our results show that the electron transmission is reduced and the electronic bandgap increased as a function of the BN concentration for different doping distribution patterns, so that (6,6)-BNC heteronanotubes become semiconducting with a tunable bandgap. The thermal conductance of helical (6,6)-BNC heteronanotubes, which is dominated by phonons, is weakly dependent on BN concentration in the range of 30-80%. Also, the Seebeck coefficient is enhanced by increasing the concentration of helical BN strips. In particular, helical (6,6)-BNC heteronanotubes with a high BN concentration (>20%) display a larger figure of merit compared to other doping distributions and, for a concentration of 50%, reach values up to 2.3 times and 3.4 times the corresponding values of a CNT at 300 K and 800 K, respectively. Our study yields new insights into the parameters tuning the thermoelectric efficiency and thus provides a starting point for designing thermoelectric devices based on BNC nanostructures.

Quantum Phonon Transport in Nanomaterials: Combining Atomistic with Non-Equilibrium Green's Function Techniques.

A crucial goal for increasing thermal energy harvesting will be to progress towards atomistic design strategies for smart nanodevices and nanomaterials. This requires the combination of computationally efficient atomistic methodologies with quantum transport based approaches. Here, we review our recent work on this problem, by presenting selected applications of the PHONON tool to the description of phonon transport in nanostructured materials. The PHONON tool is a module developed as part of the Density-Functional Tight-Binding (DFTB) software platform. We discuss the anisotropic phonon band structure of selected puckered two-dimensional materials, helical and horizontal doping effects in the phonon thermal conductivity of boron nitride-carbon heteronanotubes, phonon filtering in molecular junctions, and a novel computational methodology to investigate time-dependent phonon transport at the atomistic level. These examples illustrate the versatility of our implementation of phonon transport in combination with density functional-based methods to address specific nanoscale functionalities, thus potentially allowing for designing novel thermal devices.

Molecular and Ionic Dipole Effects on the Electronic Properties of Si-/SiO2-Grafted Alkylamine Monolayers.

In this work, we demonstrate the tunability of electronic properties of Si/SiO2 substrates by molecular and ionic surface modifications. The changes in the electronic properties such as the work function (WF) and electron affinity were experimentally measured by the contact potential difference technique and theoretically supported by density functional theory calculations. We attribute these molecular electronic effects mainly to the variations of molecular and surface dipoles of the ionic and neutral species. We have previously shown that for the alkylhalide monolayers, changing the tail group from Cl to I decreased the WF of the substrate. Here, we report on the opposite trend of WF changes, that is, the increase of the WF, obtained by using the anions of these halides from Cl- to I-. This trend was observed on self-assembled alkylammonium halide (-NH3+ X-, where X- = Cl-, Br-, or I-) monolayer-modified substrates. The monolayer's formation was supported by ellipsometry measurements, X-ray photoelectron spectroscopy, and atomic force microscopy. Comparison of the theoretical and experimental data suggests that the ionic surface dipole depends mainly on the polarizability and the position of the counter halide anion along with the organization and packaging of the layer. The described ionic modification can be easily used for facile tailoring and design of the electronic properties Si/SiO2 substrates for various device applications.

In-Situ Stretching Patterned Graphene Nanoribbons in the Transmission Electron Microscope.

The mechanical response of patterned graphene nanoribbons (GNRs) with a width less than 100 nm was studied in-situ using quantitative tensile testing in a transmission electron microscope (TEM). A high degree of crystallinity was confirmed for patterned nanoribbons before and after the in-situ experiment by selected area electron diffraction (SAED) patterns. However, the maximum local true strain of the nanoribbons was determined to be only about 3%. The simultaneously recorded low-loss electron energy loss spectrum (EELS) on the stretched nanoribbons did not reveal any bandgap opening. Density Functional Based Tight Binding (DFTB) simulation was conducted to predict a feasible bandgap opening as a function of width in GNRs at low strain. The bandgap of unstrained armchair graphene nanoribbons (AGNRs) vanished for a width of about 14.75 nm, and this critical width was reduced to 11.21 nm for a strain level of 2.2%. The measured low tensile failure strain may limit the practical capability of tuning the bandgap of patterned graphene nanostructures by strain engineering, and therefore, it should be considered in bandgap design for graphene-based electronic devices by strain engineering.