BonDNet: a graph neural network for the prediction of bond dissociation energies for charged molecules

A broad collection of technologies, including e.g. drug metabolism, biofuel combustion, photochemical decontamination of water, and interfacial passivation in energy production/storage systems rely on chemical processes that involve bond-breaking molecular reactions. In this context, a fundamental thermodynamic property of interest is the bond dissociation energy (BDE) which measures the strength of a chemical bond. Fast and accurate prediction of BDEs for arbitrary molecules would lay the groundwork for data-driven projections of complex reaction cascades and hence a deeper understanding of these critical chemical processes and, ultimately, how to reverse design them. In this paper, we propose a chemically inspired graph neural network machine learning model, BonDNet, for the rapid and accurate prediction of BDEs. BonDNet maps the difference between the molecular representations of the reactants and products to the reaction BDE. Because of the use of this difference representation and the introduction of global features, including molecular charge, it is the first machine learning model capable of predicting both homolytic and heterolytic BDEs for molecules of any charge. To test the model, we have constructed a dataset of both homolytic and heterolytic BDEs for neutral and charged (−1 and +1) molecules. BonDNet achieves a mean absolute error (MAE) of 0.022 eV for unseen test data, significantly below chemical accuracy (0.043 eV). Besides the ability to handle complex bond dissociation reactions that no previous model could consider, BonDNet distinguishes itself even in only predicting homolytic BDEs for neutral molecules; it achieves an MAE of 0.020 eV on the PubChem BDE dataset, a 20% improvement over the previous best performing model. We gain additional insight into the model''s predictions by analyzing the patterns in the features representing the molecules and the bond dissociation reactions, which are qualitatively consistent with chemical rules and intuition. BonDNet is just one application of our general approach to representing and learning chemical reactivity, and it could be easily extended to the prediction of other reaction properties in the future.

Prediction of bond dissociation energies for charged molecules with a graph neural network enabled by global molecular features and reaction difference features between products and reactants.     

A chemically consistent graph architecture for massive reaction networks applied to solid-electrolyte interphase formation

Modeling reactivity with chemical reaction networks could yield fundamental mechanistic understanding that would expedite the development of processes and technologies for energy storage, medicine, catalysis, and more. Thus far, reaction networks have been limited in size by chemically inconsistent graph representations of multi-reactant reactions (e.g. A + B → C) that cannot enforce stoichiometric constraints, precluding the use of optimized shortest-path algorithms. Here, we report a chemically consistent graph architecture that overcomes these limitations using a novel multi-reactant representation and iterative cost-solving procedure. Our approach enables the identification of all low-cost pathways to desired products in massive reaction networks containing reactions of any stoichiometry, allowing for the investigation of vastly more complex systems than previously possible. Leveraging our architecture, we construct the first ever electrochemical reaction network from first-principles thermodynamic calculations to describe the formation of the Li-ion solid electrolyte interphase (SEI), which is critical for passivation of the negative electrode. Using this network comprised of nearly 6000 species and 4.5 million reactions, we interrogate the formation of a key SEI component, lithium ethylene dicarbonate. We automatically identify previously proposed mechanisms as well as multiple novel pathways containing counter-intuitive reactions that have not, to our knowledge, been reported in the literature. We envision that our framework and data-driven methodology will facilitate efforts to engineer the composition-related properties of the SEI – or of any complex chemical process – through selective control of reactivity.

A chemically consistent graph architecture enables autonomous identification of novel solid-electrolyte interphase formation pathways from a massive reaction network.     

Interaction of Hydrogen with Ceria: Hydroxylation,Reduction, and Hydride Formation on the Surface and in the Bulk

The study reports the first attempt to address the interplay between surface and bulk in hydride formation in ceria (CeO₂) by combining experiment, using surface sensitive and bulk sensitive spectroscopic techniques on the two sample systems, i.e., CeO₂(111) thin films and CeO₂ powders, and theoretical calculations of CeO₂(111) surfaces with oxygen vacancies (O_v) at the surface and in the bulk. We show that, on a stoichiometric CeO₂(111) surface, H₂ dissociates and forms surface hydroxyls (OH). On the pre-reduced CeO_2−x samples, both films and powders, hydroxyls and hydrides (Ce−H) are formed on the surface as well as in the bulk, accompanied by the Ce³⁺ ↔ Ce⁴⁺ redox reaction. As the O_v concentration increases, hydroxyl is destabilized and hydride becomes more stable. Surface hydroxyl is more stable than bulk hydroxyl, whereas bulk hydride is more stable than surface hydride. The surface hydride formation is the kinetically favorable process at relatively low temperatures, and the resulting surface hydride may diffuse into the bulk region and be stabilized therein. At higher temperatures, surface hydroxyls can react to produce water and create additional oxygen vacancies, increasing its concentration, which controls the H₂/CeO₂ interaction. The results demonstrate a large diversity of reaction pathways, which have to be taken into account for better understanding of reactivity of ceria-based catalysts in a hydrogen-rich atmosphere.     

Supporting Middle School Teachers’ Implementation of STEM Design Challenges

We describe and analyze a professional development (PD) model that involved a partnership among science, mathematics and education university faculty, science and mathematics coordinators, and middle school administrators, teachers, and students. The overarching project goal involved the implementation of interdisciplinary STEM Design Challenges (DCs). The PD model targeted: (a) increasing teachers’ content and pedagogical content knowledge in mathematics and science; (b) helping teachers integrate STEM practices into their lessons; and (c) addressing teachers’ beliefs about engaging underperforming students in challenging problems. A unique aspect involved low‐achieving students and their teachers learning alongside each other as they co‐participated in STEM design challenges for one week in the summer. Our analysis focused on what teachers came to value about STEM DCs, and the challenges in and affordances for implementing DCs. Two significant areas of value for the teachers were students’ use of scientific, mathematical, and engineering practices and motivation, engagement, and empowerment by all learners. Challenges associated with pedagogy, curriculum, and the traditional structures of the schools were identified. Finally, there were four key affordances: (a) opportunities to construct a vision of STEM education; (b) motivation to implement DCs; (c) ambitious pedagogical tools; and, (d) ongoing support for planning and implementation. This article features a Research to Practice Companion Article . Please click on the supporting information link below to access.     

An experimentally validated rod model for soft continuum robots

Soft robots are bio-inspired, highly deformable robots with the ability to interact with workpieces in a manner that complements their hard robot counterparts. To develop practical applications and reproducible designs of soft robots, new models are necessary to describe their kinematics and dynamics. In the present work, we describe experimental and numerical investigations of a popular pneumatically-actuated soft continuum arm. These works enable us to derive constitutive relations and develop a rod model for large deformations of the arm that faithfully describes its bending behavior. We show how the resulting non-classical constitutive relation can be defined either through experiments or through quasi-static finite element simulations. With the help of this relation, the resulting rod model can be used to study the dynamics of the soft robot arm in a fast and tractable manner. (© 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)     

Modification and validation of the mixed-format Engineering Concept Assessment for middle school students using many-facet Rasch measurement

Integrating engineering into the K-12 science curriculum continues to be a focus in national reform efforts in science education. Although there is an increasing interest in research in and practice of integrating engineering in K-12 science education, to date only a few studies have focused on the development of an assessment tool to measure students’ understanding of engineering design. Most of the existing measures focus only on knowledge and understanding of engineering design concepts using multiple-choice items with the exception of the mixed-format Engineering Concept Assessment (ECA). Also, advanced measurement models are lacking application in the testing of such mixed-format assessments in science education. This study applied many-faceted Rasch measurement to the modified ECA for eighth-grade (ECA/M8) and a newly constructed rubric applied by five judges across 497 eighth-grade students’ responses after experiencing an integrated learning unit on the engineering design process. The results supported the fit of the items and rubric rating scales to the Rasch specifications. Recommendations are made for item wording, and further reliability and validity testing of the ECA/M8, and use of the ECA/M8 in science education and research.     

A forest building process on simple graphs

Consider the following process on a simple graph without isolated vertices: order the edges randomly and keep an edge if and only if it contains a vertex which is not contained in some preceding edge. The resulting set of edges forms a spanning forest of the graph.The probability of obtaining $k$ components in this process for complete bipartite graphs is determined as well as a formula for the expected number of components in any graph. A generic recurrence and some additional basic properties are discussed.     

Weak convergence of Galerkin approximations for fractional elliptic stochastic PDEs with spatial white noise

The numerical approximation of the solution to a stochastic partial differential equation with additive spatial white noise on a bounded domain is considered. The differential operator is assumed to be a fractional power of an integer order elliptic differential operator. The solution is approximated by means of a finite element discretization in space and a quadrature approximation of an integral representation of the fractional inverse from the Dunford–Taylor calculus. For the resulting approximation, a concise analysis of the weak error is performed. Specifically, for the class of twice continuously Fréchet differentiable functionals with second derivatives of polynomial growth, an explicit rate of weak convergence is derived, and it is shown that the component of the convergence rate stemming from the stochasticity is doubled compared to the corresponding strong rate. Numerical experiments for different functionals validate the theoretical results.     

Truncation of the correlation consistent basis sets: an effective approach to the reduction of computational cost?

The systematic reduction of commonly used basis sets as a means to reduce computational cost is examined for a small test set of molecules, which includes H(2), CH(4), NH(3), H(2)O, HF, and HCN. Coupled cluster with single, double, and quasiperturbative triple excitations calculations were performed using both the correlation consistent basis sets, and a set of systematically reduced basis sets to examine both the impact of the reduction upon the accuracy of the structures and energies, and the computational cost savings achieved. The effect of several truncation scenarios upon basis set convergence is also examined. Overall, for the systems studied, a reduction can occur which preserves the well-established systematic convergence behavior of the correlation consistent basis sets.     

Predicting crystal structures with data mining of quantum calculations

Predicting and characterizing the crystal structure of materials is a key problem in materials research and development. It is typically addressed with highly accurate quantum mechanical computations on a small set of candidate structures, or with empirical rules that have been extracted from a large amount of experimental information, but have limited predictive power. In this Letter, we transfer the concept of heuristic rule extraction to a large library of ab initio calculated information, and we demonstrate that this can be developed into a tool for crystal structure prediction.