Search for: All records

Forecasting models are a central part of many control systems, where high consequence decisions must be made on long latency control variables. These models are particularly relevant for emerging artificial intelligence (AI)-guided instrumentation, in which prescriptive knowledge is needed to guide autonomous decision-making. Here we describe the implementation of a long short-term memory model (LSTM) for forecasting of electron energy loss spectroscopy (EELS) data, one of the richest analytical probes of materials and chemical systems. We describe key considerations for data collection, preprocessing, training, validation, and benchmarking, showing how this approach can yield powerful predictive insight into order-disorder phase transitions. Finally, we comment on how such a model may integrate with emerging AI-guided instrumentation for powerful high-speed experimentation.

Low C-rate charge and discharge experiments, plus complementary differential voltage or differential capacity analysis, are among the most common battery characterization methods. Here, we adapt the multi-species, multi-reaction (MSMR) half-cell thermodynamic model to low C-rate cycling of whole-cell Li-ion batteries. MSMR models for the anode and cathode are coupled through whole-cell charge balances and cell-cycling voltage constraint equations, forming the basis for model-based estimation of MSMR half-cell parameters from whole-cell experimental data. Emergent properties of the whole-cell, like slippage of the anode and cathode lithiation windows, are also computed as cells cycle and degrade. A sequential least-square optimization scheme is used for parameter estimation from low-C cycling data of Samsung 18650 NMC∣C cells. Low-error fits of the open-circuit cell voltage (e.g., under 5 mV mean absolute error for charge or discharge curves) and differential voltage curves for fresh and aged cells are achieved. We explore the features (and limitations) of using literature reference values for the MSMR half-cell thermodynamic parameters (reducing our whole-cell formulation to a 1-degree-of-freedom fit) and demonstrate the benefits of expanding the degrees of freedom by letting the MSMR parameters be tailored to the cell under test, within a constrained neighborhood of the half-cell reference values. Bootstrapmore »analysis is performed on each dataset to show the robustness of our fitting to experimental noise and data sampling over the course of 600 cell cycles. The results show which specific MSMR insertion reactions are most responsible for capacity loss in each half-cell and the collective interactions that lead to whole-cell slippage and changes in useable capacity. Open-source software is made available to easily extend this model-based analysis to other labs and battery chemistries.« less

Deep eutectic solvents (DESs) are an attractive class of materials with low toxicity, broad commercial availability, low costs and simple synthesis, which allows for tuning of their properties. We develop and demonstrate the use of high-throughput and data-driven strategies to accelerate the investigation of new DES formulations. A cheminformatics approach is used to outline a design space, which results in 3477 hydrogen bond donor (HBD) and 185 quaternary ammonium salt (QAS) molecules identified as good candidate components for DES. The synthesis methodology is then adapted to a high-throughput protocol using liquid handling robots for the rapid synthesis of DES combinations. High-throughput electrochemical characterization and melting point detection systems are used to measure key performance metrics. To demonstrate the new workflow, a total of 600 unique samples are prepared and characterized, corresponding to 50 unique DES combinations at 12 HBD/QAS molar ratios. After synthesis, a total of 230 samples are found liquid at room temperature and further characterized. Several DESs display conductivities above 1 mS cm −1 , with a maximum recorded conductivity of 13.7 mS cm −1 for the combination of acetylcholine chloride (20 mol%) and ethylene glycol. All liquid DES samples show stable potential windows greater than 3 V.more »We also demonstrate that these DESs are electrochemically limited by viscosity, both in the conductivity and in the limiting processes on their cyclic voltammograms. Comparison with literature reports shows good agreement for properties measured in the high-throughput study, which helps to validate the workflow. This work demonstrates new methods to accelerate the collection of key DES metrics, providing data to formulate robust property prediction models and obtaining insight on interactions between molecular components. Data-driven high-throughput experimentation strategies can accelerate DES development for a variety of applications. Moreover, these approaches can also be extended to tackle other materials challenges with large molecular design spaces.« less

We report a comprehensive computational study of unsupervised machine learning for extraction of chemically relevant information in X-ray absorption near edge structure (XANES) and in valence-to-core X-ray emission spectra (VtC-XES) for classification of a broad ensemble of sulphorganic molecules. By progressively decreasing the constraining assumptions of the unsupervised machine learning algorithm, moving from principal component analysis (PCA) to a variational autoencoder (VAE) to t-distributed stochastic neighbour embedding (t-SNE), we find improved sensitivity to steadily more refined chemical information. Surprisingly, when embedding the ensemble of spectra in merely two dimensions, t-SNE distinguishes not just oxidation state and general sulphur bonding environment but also the aromaticity of the bonding radical group with 87% accuracy as well as identifying even finer details in electronic structure within aromatic or aliphatic sub-classes. We find that the chemical information in XANES and VtC-XES is very similar in character and content, although they unexpectedly have different sensitivity within a given molecular class. We also discuss likely benefits from further effort with unsupervised machine learning and from the interplay between supervised and unsupervised machine learning for X-ray spectroscopies. Our overall results, i.e. , the ability to reliably classify without user bias and to discover unexpected chemical signatures formore »XANES and VtC-XES, likely generalize to other systems as well as to other one-dimensional chemical spectroscopies.« less