Acceleration of Chemical Kinetics Computation with the Learned Intelligent Tabulation (LIT) Method

In this work, a data-driven methodology for modeling combustion kinetics, Learned Intelligent Tabulation (LIT), is presented. LIT aims to accelerate the tabulation of combustion mechanisms via machine learning algorithms such as Deep Neural Networks (DNNs). The high-dimensional composition space is sampled from high-fidelity simulations covering a wide range of initial conditions to train these DNNs. The input data are clustered into subspaces, while each subspace is trained with a DNN regression model targeted to a particular part of the high-dimensional composition space. This localized approach has proven to be more tractable than having a global ANN regression model, which fails to generalize across various composition spaces. The clustering is performed using an unsupervised method, Self-Organizing Map (SOM), which automatically subdivides the space. A dense network comprised of fully connected layers is considered for the regression model, while the network hyper parameters are optimized using Bayesian optimization. A nonlinear transformation of the parameters is used to improve sensitivity to minor species and enhance the prediction of ignition delay. The LIT method is employed to model the chemistry kinetics of zero-dimensional H2–O2 and CH4-air combustion. The data-driven method achieves good agreement with the benchmark method while being cheaper in terms of computational cost. LIT is naturally extensible to different combustion models such as flamelet and PDF transport models.

A Subspace-Inclusive Sampling Method for the Computational Design of Compositionally Graded Alloys

Compositionally graded alloys, a special class of functionally graded materials (FGMs), utilize localized variations in composition within a single metal part to achieve higher performance than traditional single-material parts. In previous work [1], the authors presented a computational design methodology that avoids common issues which limit a gradient alloy’s usefulness or feasibility, such as deleterious phases or properties, and also optimizes gradients for performance objectives. However, the previous methodology only samples the interior of a composition space, meaning designed gradients must include all elements in the space at every step in the gradient. Because the addition of even a small amount of an alloying element can introduce a new deleterious phase, this characteristic often neglects potentially simpler solutions to otherwise unsolvable problems and, consequently, discourages the addition of new elements to the state space. The present work improves upon the previous methodology by introducing a sampling method that includes subspaces with fewer elements in the design search. The new sampling method samples within an artificial expanded form of the state space and projects samples outside the true region to the nearest true subspace. This method is evaluated first by observing the distribution of samples in each subspace of a 2-D, 3-D, and 4-D state space. Next, a parametric study in a synthetic 2-D problem compares the performance of the new sampling scheme to the previous methodology. Lastly, the updated methodology is applied to design a gradient from stainless steel to equiatomic NiTi that has practical uses such as embedded shape memory actuation and for which the previous methodology fails to find a feasible path.

Phase constitution of the noble metal thin-film complex solid solution system Ag-Ir-Pd-Pt-Ru in dependence of elemental compositions and annealing temperatures

Multiple-principal element alloys hold great promise for multifunctional material discovery (e.g., for novel electrocatalysts based on complex solid solutions) in a virtually unlimited compositional space. Here, the phase constitution of the noble metal system Ag-Ir-Pd-Pt-Ru was investigated over a large compositional range in the quinary composition space and for different annealing temperatures from 600 to 900 °C using thin-film materials libraries. Composition-dependent X-ray diffraction mapping of the as-deposited thin-film materials library indicates different phases being present across the composition space (face-centered cubic (fcc), hexagonal close packed (hcp) and mixed fcc + hcp), which are strongly dependent on the Ru content. In general, low Ru contents promote the fcc phase, whereas high Ru contents favor the formation of an hcp solid-solution phase. Furthermore, a temperature-induced phase transformation study was carried out for a selected measurement area of fcc-Ag5Ir8Pd56Pt8Ru23. With increasing temperature, the initial fcc phase transforms to an intermediate C14-type Laves phase at 360 °C, and then to hcp when the temperature reaches 510 °C. The formation and disappearance of the hexagonal Laves phase, which covers a wide temperature range, plays a crucial role of bridging the fcc to hcp phase transition. The obtained composition, phase and temperature data are transformed into phase maps which could be used to guide theoretical studies and lay a basis for tuning the functional properties of these materials.

Data-Driven Discovery and Synthesis of High Entropy Alloy Hydrides with Targeted Thermodynamic Stability

Solid-state hydrogen storage materials that are optimized for specific use cases could be a crucial facilitator of the hydrogen economy transition. Yet the discovery of novel hydriding materials has historically been a manual process driven by chemical intuition or experimental trial-and-error. Data-driven materials' discovery paradigms provide an alternative to traditional approaches, whereby machine/statistical learning (ML) models are used to efficiently screen materials for desired properties and significantly narrow the scope of expensive/time-consuming first-principles modeling and experimental validation. Here we specifically focus on a relatively new class of hydrogen storage materials, high entropy alloy (HEA) hydrides, whose vast combinatorial composition space and local structural disorder necessitates a data-driven approach that does not rely on exact crystal structures in order to make property predictions. Our ML model quickly screens hydride stability within a large HEA space and permits down selection for laboratory validation based not only on targeted thermodynamic properties, but also secondary criteria such as alloy phase stability and density. To experimentally verify our predictions, we performed targeted synthesis and characterization of several novel hydrides that demonstrate significant destabilization (70x increase in equilibrium pressure, 20 kJ/molH₂ decrease in desorption enthalpy) relative to the benchmark HEA hydride, TiVZrNbHfH_x. Ultimately, by providing a large composition space in which hydride thermodynamics can be continuously tuned over a wide range, this work will enable efficient materials selection for various applications, especially in areas such as metal hydride based hydrogen compressors, actuators, and heat pumps.