A large scale multi-modal workflow for battery characterization: from concept to implementation

TL;DR

The paper presents a large-scale, cross-European workflow for multimodal battery characterization, aiming to correlate heterogeneous data across 15 partners and multiple facilities. It introduces a six-step, chemistry-neutral framework and two-dimensional metaviews to integrate 75 datasets and produce layered knowledge representations from individual measurements to correlative subsets. Demonstrated on Graphite/LiNiO2 full cells with/without LiTDI, the study reveals how measurement choices and observables shape interpretation and shows that distinct electrode properties can yield similar electrochemical performance. The work outlines a concrete path toward a pan-European experimental platform, emphasizing standardized protocols, FAIR data infrastructure, and ontologized tools, while acknowledging challenges in fully automated, holistic data analysis and platform scalability.

Abstract

The development of material acceleration platforms in battery research requires integrating complementary techniques and correlating heterogeneous experimental datasets. Here, this challenge is tackled in a large-scale multimodal program involving fifteen laboratories and facilities across Europe. Coordinated multi-site experiments are performed on state-of-the-art graphite / LiNiO2 Li-ion full cells to address two archetypal scientific questions: is the electrolyte composition impacting electrode properties, and how do electrode materials evolve when cells are cycled to their end-of-life? A fully standardized and centralized workflow is demonstrated, from sample production and delivery, to metadata and data handling, generating seventy-five concatenated datasets shared among all partners. Their integrated analysis shows that scientific conclusions depend critically on both the observable chosen to describe electrode properties, and the measurement technique employed. Individual experiments provide detailed information into specific aspects, such as crystal structures, redox activity, surface processes, morphology, etc., but can also function as binary diagnostic tool. Two-dimensional observable-technique patterns are introduced, in which each pixel encodes a yes, no or uncertain answer to a given scientific question. These patterns serve as multi-property metaviews, e.g. visual genotypes, enabling to classify material behavior and technique suitability according to predefined user demand and criteria, highlighting the interdependencies between measurement choices, extracted parameters and scientific interpretation. This multimodal workflow establishes a proof-of-concept for correlative analysis and underscores challenges toward fully integrated, automated and holistic approaches in energy material science.

Figures (6)

• Figure 1: Concept and practical execution of a coordinated battery investigation workflow. a) Conceptual organization. Steps 1 to 3 involve preparation, step 4 corresponds to execution and steps 5 to 6 concern data storage, sharing and results exploitation. The analyses in step 6 can be divided into three categories: single dataset exploitation, global multi-property mapping, and correlative analysis. The light gray boxes and orange arrows indicate who is involved in each step and what specific actions are performed. b) Practical implementation of the workflow to study Gr / LNO batteries. Battery-making expert partners are indicated in green, characterization expert partners in purple. Time-sequenced execution of the 6 workflow steps, indicating who is involved (green / purple persons), what decisions are made and actions taken (gray boxes), and what are the main quantified outcomes (bottom text): 1) two scientific questions defined and battery components selected, including electrolyte 1 (standard) and electrolyte 2 (with additive); 2) selection of 15 relevant techniques; 3) production and shipping of 90 samples; 4) execution of 19 experiments; 5) 75 datasets and metadata handling through a common cloud infrastructure and 6) data analysis using three-layers: 75 individual datasets; 4 global multi-property metaviews and some correlatively-analyzed subsets. The timescale of the last analysis step has been limited to 3 months in our proof-of-concept application, but can be extended depending on the depth of targeted sub-analyses.
• Figure 2: Techniques, observables and samples. a) Detailed view of workflow step 2 showing the combination of 15 techniques employed (colored boxes) and the physico-chemical parameters extracted (gray lines) to analyze crystal and electronic structures, phase transformation, morphology, Li distribution, reaction dynamics, homogeneity and SEI and surface properties. The experiments where performed on centrally standardized samples obtained from coin cell cycling. b) Execution of step 3 comprising the standardization of sample preparation and sample exchange between partners. Galvanostatic cycling data (discharge capacity vs. cycle number) of 35 coin cells assembled with the two electrolyte types, showing 1C charge / discharge sequences followed by control measurement at C/20 to regularly evaluate the remaining capacity. The end-of-life is represented by the dashed lines, and corresponds to the cells being dismounted in discharged state to prepare the "aged electrodes", e.g. aged LNO in lithiated state and aged graphite in delithiated state. Electrolyte with LiTDI is plotted in green, and electrolyte without LiTDI in pink.
• Figure 3: Isolated dataset and result classification using global multi-property 2D metaviews. a) Stand-alone result: X-ray nano holotomography of LNO electrodes. 3D reconstructed volumes are shown for pristine as well as aged with and without LiTDI samples at high resolution with the corresponding estimated porosity ($\upphi$). The technique enables to track morphological evolutions during cycling with the two electrolytes (SQ1 and SQ2). b) LNO metaviews. c) Graphite metaviews. A metaview is a (15, 8) matrix (15 techniques, 8 observables) reporting discrete answers to SQ1 / SQ2 (middle and right panels, respectively) into colored pixels (yes = green; no = orange; unsure = yellow; non applicable = gray; not done = uncolored). SQ1: "Is there a difference between samples prepared with different electrolytes?" and SQ2: "Is there a difference between pristine and EoL samples?". Global views (left panel) are composed by averaging the parameter-specific answers using a prioritized logical rule sequence: i) one green = green; ii) one yellow = yellow; iii) one orange = orange; etc.. Dashed rectangles highligth specific examples of correlated subset of results, picking one given technique (red applied to LNO for SQ1 and SQ2) or one particular parameter across all techniques (green for graphite, purple for LNO, applied to SQ1).
• Figure 4: Correlative analysis in subareas of the metaviews. a) Correlative sublayer: SQ1, LNO, Li distribution, probed by Neutron Depth Profiling (NDP, left panel), Neutron Imaging (NI, middle panel), and lab-scale Raman mapping (right panel). In line "Raman mapping" and column "Li distribution”, the pixel is orange, as for line "NI", meaning that no noticeable differences in Li distribution have been obtained using these techniques when comparing LNO electrodes aged with both electrolytes, as seen from the similar attenuated neutron intensities (in red and blue with and without LiTDI respectively) and similar Raman peaks positions and shapes (red and blue spectra). In contrast, NDP revealed changes in the normalized lithium concentration measured in the depth (red and blue data), hence the pixel is colored green. b) Correlative sublayer: SQ1, Graphite, SEI and interfaces investigated with Nuclear Magnetic Resonance (NMR, left panel), lab-scale X-ray Photoelectron Spectroscopy (XPS, middle panel) and synchrotron X-ray Raman Spectroscopy (XRS, right panel). The electrolyte impact on the SEI nature is probed with different sensitivities. In lines "NMR" and "XPS", the pixel is green as these techniques evidenced differences in the SEI structure and composition in between graphite electrodes cycled with both electrolytes, seen from changes in the $^{7}$Li and $^{19}$F spectra as well as large-energy band XPS survey spectra. The cell for the "XRS synchrotron" line is orange as no variations were detected with enough intensity with regards to the measurement error margins, as seen from the O edge data.
• Figure 5: Roadmap towards fully holistic characterization workflows. a) Stages of workflow complexity. Proof-of-concept of Stage III ("Correlated techniques") is reported in this work. Stage IV necessitates to overcome challenges (b) and implement an array of tools (c). Challenges encompass communication, organization, standardization and data correlation aspects. A Battery Characterization Platform toolbox should enable users to conceive and execute workflows, with easy access to input / output knowledge and data. Workflow builder, databases, data infrastructures and ontology are coordinated via a centralized communication hub.