Deep Learning Spectroscopy: Neural Networks for Molecular Excitation Spectra

Deep learning methods for the prediction of molecular excitation spectra are presented. For the example of the electronic density of states of 132k organic molecules, three different neural network architectures: multilayer perceptron (MLP), convolutional neural network (CNN), and deep tensor neural network (DTNN) are trained and assessed. The inputs for the neural networks are the coordinates and charges of the constituent atoms of each molecule. Already, the MLP is able to learn spectra, but the root mean square error (RMSE) is still as high as 0.3 eV. The learning quality improves significantly for the CNN (RMSE = 0.23 eV) and reaches its best performance for the DTNN (RMSE = 0.19 eV). Both CNN and DTNN capture even small nuances in the spectral shape. In a showcase application of this method, the structures of 10k previously unseen organic molecules are scanned and instant spectra predictions are obtained to identify molecules for potential applications.

Submolecular Imaging by Noncontact Atomic Force Microscopy with an Oxygen Atom Rigidly Connected to a Metallic Probe

In scanning probe microscopy, the imaging characteristics in the various interaction channels crucially depend on the chemical termination of the probe tip. Here we analyze the contrast signatures of an oxygen-terminated copper tip with a tetrahedral configuration of the covalently bound terminal O atom. Supported by first-principles calculations we show how this tip termination can be identified by contrast analysis in noncontact atomic force and scanning tunneling microscopy (NC-AFM, STM) on a partially oxidized Cu(110) surface. After controlled tip functionalization by soft indentations of only a few angstroms in an oxide nanodomain, we demonstrate that this tip allows imaging an organic molecule adsorbed on Cu(110) by constant-height NC-AFM in the repulsive force regime, revealing its internal bond structure. In established tip functionalization approaches where, for example, CO or Xe is deliberately picked up from a surface, these probe particles are only weakly bound to the metallic tip, leading to lateral deflections during scanning. Therefore, the contrast mechanism is subject to image distortions, artifacts, and related controversies. In contrast, our simulations for the O-terminated Cu tip show that lateral deflections of the terminating O atom are negligible. This allows a detailed discussion of the fundamental imaging mechanisms in high-resolution NC-AFM experiments. With its structural rigidity, its chemically passivated state, and a high electron density at the apex, we identify the main characteristics of the O-terminated Cu tip, making it a highly attractive complementary probe for the characterization of organic nanostructures on surfaces.

Atomic structures and orbital energies of 61,489 crystal-forming organic molecules

Data science and machine learning in materials science require large datasets of technologically relevant molecules or materials. Currently, publicly available molecular datasets with realistic molecular geometries and spectral properties are rare. We here supply a diverse benchmark spectroscopy dataset of 61,489 molecules extracted from organic crystals in the Cambridge Structural Database (CSD), denoted OE62. Molecular equilibrium geometries are reported at the Perdew-Burke-Ernzerhof (PBE) level of density functional theory (DFT) including van der Waals corrections for all 62 k molecules. For these geometries, OE62 supplies total energies and orbital eigenvalues at the PBE and the PBE hybrid (PBE0) functional level of DFT for all 62 k molecules in vacuum as well as at the PBE0 level for a subset of 30,876 molecules in (implicit) water. For 5,239 molecules in vacuum, the dataset provides quasiparticle energies computed with many-body perturbation theory in the G0W0 approximation with a PBE0 starting point (denoted GW5000 in analogy to the GW100 benchmark set (M. van Setten et al. J. Chem. Theory Comput. 12, 5076 (2016))).

Understanding scanning tunneling microscopy contrast mechanisms on metal oxides: a case study

A comprehensive analysis of contrast formation mechanisms in scanning tunneling microscopy (STM) experiments on a metal oxide surface is presented with the oxygen-induced (2√2×√2)R45° missing row reconstruction of the Cu(100) surface as a model system. Density functional theory and electronic transport calculations were combined to simulate the STM imaging behavior of pure and oxygen-contaminated metal tips with structurally and chemically different apexes while systematically varying bias voltage and tip-sample distance. The resulting multiparameter database of computed images was used to conduct an extensive comparison with experimental data. Excellent agreement was attained for a large number of cases, suggesting that the assumed model tips reproduce most of the commonly encountered contrast-determining effects. Specifically, we find that depending on the bias voltage polarity, copper-terminated tips allow selective imaging of two structurally distinct surface Cu sites, while oxygen-terminated tips show complex contrasts with pronounced asymmetry and tip-sample distance dependence. Considering the structural and chemical stability of the tips reveals that the copper-terminated apexes tend to react with surface oxygen at small tip-sample distances. In contrast, oxygen-terminated tips are considerably more stable, allowing exclusive surface oxygen imaging at small tip-sample distances. Our results provide a conclusive understanding of fundamental STM imaging mechanisms, thereby providing guidelines for experimentalists to achieve chemically selective imaging by properly selecting imaging parameters.

Efficient Amino Acid Conformer Search with Bayesian Optimization

Finding low-energy molecular conformers is challenging due to the high dimensionality of the search space and the computational cost of accurate quantum chemical methods for determining conformer structures and energies. Here, we combine active-learning Bayesian optimization (BO) algorithms with quantum chemistry methods to address this challenge. Using cysteine as an example, we show that our procedure is both efficient and accurate. After only 1000 single-point calculations and approximately 80 structure relaxations, which is less than 10% computational cost of the current fastest method, we have found the low-energy conformers in good agreement with experimental measurements and reference calculations. To test the transferability of our method, we also repeated the conformer search of serine, tryptophan, and aspartic acid. The results agree well with previous conformer search studies.

Detecting stable adsorbates of (1S)-camphor on Cu(111) with Bayesian optimization

Identifying the atomic structure of organic-inorganic interfaces is challenging with current research tools. Interpreting the structure of complex molecular adsorbates from microscopy images can be difficult, and using atomistic simulations to find the most stable structures is limited to partial exploration of the potential energy surface due to the high-dimensional phase space. In this study, we present the recently developed Bayesian Optimization Structure Search (BOSS) method as an efficient solution for identifying the structure of non-planar adsorbates. We apply BOSS with density-functional theory simulations to detect the stable adsorbate structures of (1S)-camphor on the Cu(111) surface. We identify the optimal structure among eight unique types of stable adsorbates, in which camphor chemisorbs via oxygen (global minimum) or physisorbs via hydrocarbons to the Cu(111) surface. This study demonstrates that new cross-disciplinary tools, such as BOSS, facilitate the description of complex surface structures and their properties, and ultimately allow us to tune the functionality of advanced materials.

Updates to the DScribe library: New descriptors and derivatives

We present an update of the DScribe package, a Python library for atomistic descriptors. The update extends DScribe's descriptor selection with the Valle-Oganov materials fingerprint and provides descriptor derivatives to enable more advanced machine learning tasks, such as force prediction and structure optimization. For all descriptors, numeric derivatives are now available in DScribe. For the many-body tensor representation (MBTR) and the Smooth Overlap of Atomic Positions (SOAP), we have also implemented analytic derivatives. We demonstrate the effectiveness of the descriptor derivatives for machine learning models of Cu clusters and perovskite alloys.

Density-functional theory study of gramicidin A ion channel geometry and electronic properties

Understanding the mechanisms underlying ion channel function from the atomic-scale requires accurate ab initio modelling as well as careful experiments. Here, we present a density functional theory (DFT) study of the ion channel gramicidin A (gA), whose inner pore conducts only monovalent cations and whose conductance has been shown to depend on the side chains of the amino acids in the channel. We investigate the ground state geometry and electronic properties of the channel in vacuum, focusing on their dependence on the side chains of the amino acids. We find that the side chains affect the ground state geometry, while the electrostatic potential of the pore is independent of the side chains. This study is also in preparation for a full, linear scaling DFT study of gA in a lipid bilayer with surrounding water. We demonstrate that linear scaling DFT methods can accurately model the system with reasonable computational cost. Linear scaling DFT allows ab initio calculations with 10,000-100,000 atoms and beyond, and will be an important new tool for biomolecular simulations.

Efficient hyperparameter tuning for kernel ridge regression with Bayesian optimization

Machine learning methods usually depend on internal parameters—so called hyperparameters—that need to be optimized for best performance. Such optimization poses a burden on machine learning practitioners, requiring expert knowledge, intuition or computationally demanding brute-force parameter searches. We here assess three different hyperparameter selection methods: grid search, random search and an efficient automated optimization technique based on Bayesian optimization (BO). We apply these methods to a machine learning problem based on kernel ridge regression in computational chemistry. Two different descriptors are employed to represent the atomic structure of organic molecules, one of which introduces its own set of hyperparameters to the method. We identify optimal hyperparameter configurations and infer entire prediction error landscapes in hyperparameter space that serve as visual guides for the hyperparameter performance. We further demonstrate that for an increasing number of hyperparameters, BO and random search become significantly more efficient in computational time than an exhaustive grid search, while delivering an equivalent or even better accuracy.

Exploring the Conformers of an Organic Molecule on a Metal Cluster with Bayesian Optimization

Finding low-energy conformers of organic molecules is a complex problem due to the flexibilities of the molecules and the high dimensionality of the search space. When such molecules are on nanoclusters, the search complexity is exacerbated by constraints imposed by the presence of the cluster and other surrounding molecules. To address this challenge, we modified our previously developed active learning molecular conformer search method based on Bayesian optimization and density functional theory. Especially, we have developed and tested strategies to avoid steric clashes between a molecule and a cluster. In this work, we chose a cysteine molecule on a well-studied gold-thiolate cluster as a model system to test and demonstrate our method. We found that cysteine conformers in a cluster inherit the hydrogen bond types from isolated conformers. However, the energy rankings and spacings between the conformers are reordered.

Elevated serum dipeptidyl peptidase IV activity in patients with chronic tonsillitis

Background: Palatine tonsils represent the first place of contact for a variety of antigenic substances present in air and food. Upon antigen stimulation, the interactions between T and B lymphocytes in the tonsil are known to depend on the expression of different co-stimulatory molecules, including proteolytic ectoenzymes. Dipeptidyl peptidase IV (DPP IV) and aminopeptidase N (APN), as T lymphocyte co-stimulatory molecules, participate in the regulation of the immune response during inflammation.

Methods: In this study, the serum and lymphocyte enzymatic activity of DPP IV and APN was investigated in 32 patients, 13 with recurrent tonsillitis (RT) and 19 with tonsillar hypertrophy (TH), before and one month after tonsillectomy. The enzymatic activity of DPP IV and APN in tonsillar lymphocytes and serum was determined kinetically at 37°C using Gly-Pro- p-nitroanilide (for DPP IV) and Ala- p-nitroanilide (for APN) as chromogenic substrates.

Results: Significantly higher serum DPP IV and APN activities ( P<0.001) were found in TH patients compared with those with RT before tonsillectomy. DPP IV activity in TH patients was also elevated compared with the control of the same age ( P<0.001), whereas the activity of APN was the same as the control group. The activity of both enzymes was the same as of controls after tonsillectomy. In addition, the results show that DPP IV and APN activities in serum decrease significantly with age. Tonsillar lymphocytes demonstrated a wide range of DPP IV and APN activities without significant differences between the investigated groups.

Conclusion: An increased serum DPP IV activity was observed in TH patients compared with both RT patients and controls before tonsillectomy. After tonsillectomy, all activities were similar. The results suggest that serum DPP IV activity may have potential as a diagnostic marker for patients with TH.

First Principles Calculations on the Stoichiometric and Defective (101) Anatase Surface and Upon Hydrogen and H2Pc Adsorption: The Influence of Electronic Exchange and Correlation and of Basis Set Approximations

Anatase TiO₂ provides photoactivity with high chemical stability at a reasonable cost. Different methods have been used to enhance its photocatalytic activity by creating band gap states through the introduction of oxygen vacancies, hydrogen impurities, or the adorption of phthalocyanines, which are usually employed as organic dyes in dye-sensitized solar cells. Predicting how these interactions affect the electronic structure of anatase requires an efficient and robust theory. In order to document the efficiency and accuracy of commonly used approaches we have considered two widely used implementations of density functional theory (DFT), namely the all-electron linear combination of atomic orbitals (AE–LCAO) and the pseudo-potential plane waves (PP–PW) approaches, to calculate the properties of the stoichiometric and defective anatase TiO₂ (101) surface. Hybrid functionals, and in particular HSE, lead to a computed band gap in agreement with that measured by using UV adsorption spectroscopy. When using PBE+U, the gap is underestimated by 20 % but the computed position of defect induced gap states relative to the conduction band minimum (CBM) are found to be in good agreement with those calculated using hybrid functionals. These results allow us to conclude that hybrid functionals based on the use of AE–LCAO provide an efficient and robust approach for predicting trends in the band gap and the position of gap states in large model systems. We extend this analysis to surface adsorption and use the AE–LCAO approach with the hybrid functional HSED3 to study the adsorption of the phthalocyanine H₂Pc on anatase (101). Our results suggest that H₂Pc prefers to be adsorbed on the surface Ti_5c rows of anatase (101), in agreement with that seen in recent STM experiments on rutile (110).

Arterial ring encircling the bile duct with coexisting anomalous liver irrigation

Analysis of the arterial system of the liver in a 64-year-old male cadaver revealed an anomaly of the branching of the proper hepatic artery. The right branch of the hepatic artery was missing, replaced by two segmental arteries, so that a 'trifurcation' occurred, consisting of two segmental arteries, anterior segmental and posterior segmental, and the left branch of the hepatic artery. These segmental arteries formed an arterial ring around the common hepatic duct, completely encircling it. A single cystic artery unusually originated from the posterior segmental artery. An extremely rare anatomical variation of the dorsal liver segment irrigation proves the fact that anatomical anomalies are usually multiple. The arterial blood supply for the dorsal liver segment is through four separate vessels ('quadruple type'), two originating from the left branch of the hepatic artery and two from the posterior segmental artery. These four arteries presented numerous anastomotic bridges among themselves.