Computational methods that automatically extract knowledge from data are critical for enabling data-driven materials science. A reliable identification of lattice symmetry is a crucial first step for ...materials characterization and analytics. Current methods require a user-specified threshold, and are unable to detect average symmetries for defective structures. Here, we propose a machine learning-based approach to automatically classify structures by crystal symmetry. First, we represent crystals by calculating a diffraction image, then construct a deep learning neural network model for classification. Our approach is able to correctly classify a dataset comprising more than 100,000 simulated crystal structures, including heavily defective ones. The internal operations of the neural network are unraveled through attentive response maps, demonstrating that it uses the same landmarks a materials scientist would use, although never explicitly instructed to do so. Our study paves the way for crystal structure recognition of-possibly noisy and incomplete-three-dimensional structural data in big-data materials science.
Abstract
Due to their ability to recognize complex patterns, neural networks can drive a paradigm shift in the analysis of materials science data. Here, we introduce ARISE, a crystal-structure ...identification method based on Bayesian deep learning. As a major step forward, ARISE is robust to structural noise and can treat more than 100 crystal structures, a number that can be extended on demand. While being trained on ideal structures only, ARISE correctly characterizes strongly perturbed single- and polycrystalline systems, from both synthetic and experimental resources. The probabilistic nature of the Bayesian-deep-learning model allows to obtain principled uncertainty estimates, which are found to be correlated with crystalline order of metallic nanoparticles in electron tomography experiments. Applying unsupervised learning to the internal neural-network representations reveals grain boundaries and (unapparent) structural regions sharing easily interpretable geometrical properties. This work enables the hitherto hindered analysis of noisy atomic structural data from computations or experiments.
Ultrathin black phosphorus is a two-dimensional semiconductor with a sizeable band gap. Its excellent electronic properties make it attractive for applications in transistor, logic and optoelectronic ...devices. However, it is also the first widely investigated two-dimensional material to undergo degradation upon exposure to ambient air. Therefore a passivation method is required to study the intrinsic material properties, understand how oxidation affects the physical properties and enable applications of phosphorene. Here we demonstrate that atomically thin graphene and hexagonal boron nitride can be used for passivation of ultrathin black phosphorus. We report that few-layer pristine black phosphorus channels passivated in an inert gas environment, without any prior exposure to air, exhibit greatly improved n-type charge transport resulting in symmetric electron and hole transconductance characteristics.
We demonstrate a straightforward and effective laser pruning approach to reduce multilayer black phosphorus (BP) to few-layer BP under ambient condition. Phosphorene oxides and suboxides are formed ...and the degree of laser-induced oxidation is controlled by the laser power. Since the band gaps of the phosphorene suboxide depend on the oxygen concentration, this simple technique is able to realize localized band gap engineering of the thin BP. Micropatterns of few-layer phosphorene suboxide flakes with unique optical and fluorescence properties are created. Remarkably, some of these suboxide flakes display long-term (up to 2 weeks) stability in ambient condition. Comparing against the optical properties predicted by first-principle calculations, we develop a “calibration” map in using focused laser power as a handle to tune the band gap of the BP suboxide flake. Moreover, the surface of the laser patterned region is altered to be sensitive to toxic gas by way of fluorescence contrast. Therefore, the multicolored display is further demonstrated as a toxic gas monitor. In addition, the BP suboxide flake is demonstrated to exhibit higher drain current modulation and mobility comparable to that of the pristine BP in the electronic application.
Display omitted
•We model two P3HT and singled-walled carbon nanotube interfaces.•We examine charge redistribution at the two interfaces.•We determine donor–acceptor coupling affects charge transfer ...amount and work functions dominate charge transfer direction.
Electronic structures and charge redistribution between P3HT and (10,2)/(6,5) carbon nanotubes (CNT) are investigated by density functional theory calculations. The simulations show that electron is transferred from flat P3HT to (10,2) CNT while hole is transferred from (6,5) CNT to wrapped P3HT due to the different work functions of these materials. The large built-in potential can compete to exciton binding energy, leading to efficient charge separation across the type-II photovoltaic heterojunctions. Electron transfer faster than hole is expected because the electron donor state is much more delocalization, creating larger donor–acceptor coupling, which provides critical insights of organic photovoltaic solar cells.
Abstract
A public data-analytics competition was organized by the Novel Materials Discovery (NOMAD) Centre of Excellence and hosted by the online platform Kaggle by using a dataset of 3,000 (Al
x
Ga
...y
In
1–
x
–
y
)
2
O
3
compounds. Its aim was to identify the best machine-learning (ML) model for the prediction of two key physical properties that are relevant for optoelectronic applications: the electronic bandgap energy and the crystalline formation energy. Here, we present a summary of the top-three ranked ML approaches. The first-place solution was based on a crystal-graph representation that is novel for the ML of properties of materials. The second-place model combined many candidate descriptors from a set of compositional, atomic-environment-based, and average structural properties with the light gradient-boosting machine regression model. The third-place model employed the smooth overlap of atomic position representation with a neural network. The Pearson correlation among the prediction errors of nine ML models (obtained by combining the top-three ranked representations with all three employed regression models) was examined by using the Pearson correlation to gain insight into whether the representation or the regression model determines the overall model performance. Ensembling relatively decorrelated models (based on the Pearson correlation) leads to an even higher prediction accuracy.
Electronic health records (EHR) and claims data are rich sources of real-world data that reflect patient health status and healthcare utilization. Querying these databases to answer epidemiological ...questions is challenging due to the intricacy of medical terminology and the need for complex SQL queries. Here, we introduce an end-to-end methodology that combines text-to-SQL generation with retrieval augmented generation (RAG) to answer epidemiological questions using EHR and claims data. We show that our approach, which integrates a medical coding step into the text-to-SQL process, significantly improves the performance over simple prompting. Our findings indicate that although current language models are not yet sufficiently accurate for unsupervised use, RAG offers a promising direction for improving their capabilities, as shown in a realistic industry setting.
As a result of the computing power provided by the current technology, computational methods now play an important role in modeling and designing materials at the nanoscale. The focus of this ...dissertation is two-fold: first, new computational methods to model nanoscale transport are introduced, then state-of-the-art tools based on density functional theory are employed to explore the properties of phosphorene, a novel low dimensional material with great potential for applications in nanotechnology. A Wannier function description of the electron density is combined with a generalized Slater-Koster interpolation technique, enabling the introduction of a new computational method for constructing first-principles model Hamiltonians for electron and hole transport that maintain the density functional theory accuracy at a fraction of the computational cost. As a proof of concept, this new approach is applied to model polythiophene, a polymer ubiquitous in organic photovoltaic devices. A new low dimensional material, phosphorene - a single layer of black phosphorous - the phosphorous analogue of graphene was first isolated in early 2014 and has attracted considerable attention. It is a semiconductor with a sizable band gap, which makes it a perfect candidate for ultrathin transistors. Multi-layer phosphorene transistors have already achieved the highest hole mobility of any two-dimensional material apart from graphene. Phosphorene is prone to oxidation, which can lead to degradation of electrical properties, and eventually structural breakdown. The calculations reported here are some of the first to explore this oxidation and reveal that different types of oxygen defects are readily introduced in the phosphorene lattice, creating electron traps in some situations. These traps are responsible for the non-ambipolar behavior observed by experimental collaborators in air-exposed few-layer black phosphorus devices. Calculation results predict that air exposure of phosphorene creates a new family of two-dimensional oxides, which has been later confirmed by X-ray photoemission measurements. These oxides can form protective coatings for phosphorene and have interesting tunable electronic properties. Finally, Wannier function interpolation has been used to demonstrate that a saddle-point van Hove singularity is present near the phosphorene Fermi energy, as observed in some layered cuprate high temperature superconductors; this leads to an intriguing strain-induced ferromagnetic instability.
Due to their ability to recognize complex patterns, neural networks can drive a paradigm shift in the analysis of materials science data. Here, we introduce ARISE, a crystal-structure identification ...method based on Bayesian deep learning. As a major step forward, ARISE is robust to structural noise and can treat more than 100 crystal structures, a number that can be extended on demand. While being trained on ideal structures only, ARISE correctly characterizes strongly perturbed single- and polycrystalline systems, from both synthetic and experimental resources. The probabilistic nature of the Bayesian-deep-learning model allows to obtain principled uncertainty estimates, which are found to be correlated with crystalline order of metallic nanoparticles in electron tomography experiments. Applying unsupervised learning to the internal neural-network representations reveals grain boundaries and (unapparent) structural regions sharing easily interpretable geometrical properties. This work enables the hitherto hindered analysis of noisy atomic structural data from computations or experiments.