Arnulf Latz’s research while affiliated with Ulm University and other places

In-orbit satellite REIMEI, developed by the Japan Aerospace Exploration Agency (JAXA), has been relying on off-the-shelf Li-ion batteries since its launch in 2005 [1]. The data from the batteries have been recorded and analyzed ever since. These comprise current and voltage measurements of several thousand cycles [2,3]. The performance and durability of Li-ion batteries is impacted by various degradation mechanisms, one of which is the growth of the solid-electrolyte interphase (SEI). Long-term SEI growth is the greatest contributor to capacity fade in lithium-ion batteries. Physics-based models for long-term SEI growth have been developed [4,5]. To show the inhomogeneous growth of the SEI in 3D, we performed microstructure-resolved simulations [6]. In this contribution, we will address several aspects of the analysis and simulation of the batteries of satellite REIMEI. We simulate long-term degradation under the generic LEO satellite cycling conditions in a P2D framework. The simulations are validated with terrestrial experiments and in-flight data provided by JAXA [1,6]. These studies are the foundation for analyzing the states of the batteries, which cannot be measured directly. To estimate the state of charge and state of health, we make use of filter techniques and the in-flight data of the satellite batteries. Kalman filters are particularly suitable for the noisy data. Since the states change on different timescales, a multi-time-scale algorithm is developed, where two filters are combined to estimate the states simultaneously. With this approach, we aim to reliably predict the lifetime of satellite batteries in orbit [3]. References [1] M. Uno, et al. , J. Power Sources , 196(20) (2011) 8755–8763. [2] O. Mendoza-Hernandez, et al. , Electrochemistry 88 ( 4 ) (2020) 300-304. [3] L. Bolay, PhD thesis Ulm University , (2024), https://doi.org/10.18725/OPARU-52298 [4] F. Single, et al. , ChemSusChem , 11(12) (2018) 1950–1955. [5] L. von Kolzenberg, et al. , ChemSusChem , 13(15) (2020) 3901–3910. [6] L. Bolay, et al. , J. Power Sources Advances , 14 (2022) 100083.

Silicon is widely considered to be a promising next-generation anode material, primarily due to its remarkably high theoretical capacity. Furthermore, silicon is an abundant, cheap, and widely spread material. However, a major challenge for the commercialization of silicon anodes is the significant voltage hysteresis reducing efficiency and leading to detrimental heat generation during fast-charging. Additionally, the hysteresis causes an unclear state-of-charge (SOC) to voltage relation impeding precise SOC estimation. The voltage hysteresis behavior of silicon anodes is addressed in literature with three different arguments: phase transformations, plastic flow of silicon, and slow diffusion. These approaches can interpret the voltage hysteresis of crystalline silicon, thin films, and large anode particles, respectively. Nevertheless, we show that they cannot explain the hysteresis observed for silicon anodes consisting of amorphous nanoparticles, the relevant material for next-generation lithium-ion batteries. Our investigation highlights the chemo-mechanical interplay between the silicon nanoparticle and the covering Solid-Electrolyte Interphase (SEI) as the underlying cause for the significant voltage hysteresis. The SEI, a thin passivating layer, forms on negative electrode particles due to electrolyte decomposition [1]. In the case of silicon particles, the volume changes during lithiation and delithiation result in massive strains and plastic deformation occurring within the SEI [2]. Our chemo-mechanical particle-SEI description successfully replicates the observed open-circuit voltage hysteresis in experiments and aligns with the Plett model [3]. Moreover, our visco-elastoplastic SEI model reproduces the voltage difference between slow cycling and the relaxed open-circuit voltage. In our recent work, we show that a sophisticated mechanical model explains the slow voltage relaxation observed for silicon nanoparticles as well as the observed C rate dependence [4]. To conclude, we explain the voltage hysteresis and the slow voltage relaxation of silicon nanoparticles with a visco-elastoplastic particle-SEI model and discuss options to mitigate the size of the voltage hysteresis. This detailed physical understanding can improve the performance of all-silicon anodes and contribute to their commercialization. [1] Köbbing, L.; Latz, A.; Horstmann, B. J. Power Sources 2023 , DOI: 10.1016/j.jpowsour.2023.232651. [2] Kolzenberg, L.; Latz, A.; Horstmann, B. Batter. Supercaps 2022 , DOI: 10.1002/batt.202100216. [3] Köbbing, L.; Latz, A.; Horstmann, B. Adv. Funct. Mater . 2024 , DOI: 10.1002/adfm.202308818. [4] Köbbing, L.; Latz, A.; Horstmann, B. (in preparation) . Figure 1

Nickel‐rich stoichiometries such as NMC811 have gained increasing relevance for lithium‐ion‐batteries in recent years due to their high specific capacity and reduced use of critical resources. However, low intrinsic electronic conductivity of NMC active materials makes the use of carbon‐based additives necessary. Volume fraction and distribution of the carbon‐binder‐domain (CBD) have a significant impact on the electrode performance. This work combines high‐resolution tomography and microstructure‐resolved simulations to characterize the three‐dimensional transport networks of a commercial NMC811 cathode. FIB‐SEM tomography reveals that low CBD volume fractions with suboptimal distribution cause a non‐percolating conductive network in the microstructure and thus unfavourably low electronic conductivity. Increasing the CBD content through virtual electrode design enables percolation and enhances electronic conductivity fundamentally. Simulations on both the real and virtually designed structures demonstrate how percolating CBD networks lead to a significantly improved energy density.

To further improve Lithium-ion batteries (LiBs), a profound understanding of complex battery processes is crucial. Physical models offer understanding but are difficult to validate and parameterize. Therefore, automated machine-learning methods (ML) are necessary to evaluate models with experimental data. Bayesian methods, e.g., Bayesian optimization for likelihood-free inference (EP-BOLFI), stand out as they capture uncertainties in models and data while granting meaningful parameterization. An important topic is prolonging battery lifetime, which is limited by degradation, such as the solid-electrolyte interphase (SEI) growth. As a case study, we apply EP-BOLFI to parametrize SEI growth models with synthetic and real degradation data. EP-BOLFI allows for incorporating human expertise in the form of suitable feature selection, which improves the parametrization. We show that even under impeded conditions, we achieve correct parameterization with reasonable uncertainty quantification, needing less computational effort than standard Markov chain Monte Carlo methods. Additionally, the physically reliable summary statistics show if parameters are strongly correlated and not unambiguously identifiable. Further, we investigate Bayesian alternately subsampled quadrature (BASQ), which calculates model probabilities, to confirm electron diffusion as the best theoretical model to describe SEI growth during battery storage.

The conductive additive and binder domain (CBD) is an essential component of lithium-ion battery electrodes. It enhances the electrical connectivity and mechanical stability within the solid electrode matrix. Migration of the binder during electrode drying can lead to an inhomogeneous distribution of the CBD, impeding transport of lithium ions into the electrodes, and diminishing the electronic pathways between solid particles and the current collector. This is especially prominent in thick electrodes at high drying rates. Therefore, we investigate the effect of a non-uniform CBD distribution on the electrochemical performance of NMC622 electrodes via microstructure-resolved three-dimensional (3D) simulations on virtual electrodes, based on tomographic image data, and compare them with experimental results. The valuable information derived by combining microstructure-resolved models with electrochemical impedance spectroscopy measurements on symmetric cells under blocking electrolyte conditions is used to characterize the lithium-ion transport in the electrode pore space, including the contributions of the CBD. The effect of this inhomogeneity on electrode performance is then gauged via galvanostatic discharge simulations under changing discharge currents and for varying electrode densities. Through our work, we demonstrate the significance of the CBD distribution and enable predictive simulations for future battery design.

This work compares a state of the art data‐driven model to predict the state of health (SoH) in lithium ion batteries with a new prediction model based on the mechanistic framework. The mechanistic approach attributes the degradation to individual components such as loss of available capacity on each electrode as well as loss of cyclable lithium. By combining the mechanistic framework with data‐driven models for the component losses based on a design of experiment, we achieve a cycle aging model that can predict capacity degradation as well as degradation‐induced changes to the discharge potential curve. Using this cycle aging model alongside with a semi‐empirical calendar aging model, we present a holistic aging model that we validate on independent validation tests containing time‐variant load profiles. While the purely data‐driven model is better at predicting the SoH, the mechanistic model clearly has it advantages in a deeper understanding that can potentially enhance the current methods of tracking and updating the characteristic open‐circuit voltage curve over lifetime.

Understanding the mechanical interplay between silicon anodes and their surrounding solid-electrolyte interphase (SEI) is essential to improve the next generation of lithium-ion batteries. We model and simulate a 2D elliptical silicon nanowire with SEI via a thermodynamically consistent chemo-mechanical continuum ansatz using a higher order finite element method in combination with a variable-step, variable-order time integration scheme. Considering a soft viscoplastic SEI for three half cycles, we see at the minor half-axis the largest stress magnitude at the silicon nanowire surface, leading to a concentration anomaly. This anomaly is caused by the shape of the nanowire itself and not by the SEI. Also for the tangential stress of the SEI, the largest stress magnitudes are at this point, which can lead to SEI fracture. However, for a stiff SEI, the largest stress magnitude inside the nanowire occurs at the major half-axis, causing a reduced concentration distribution in this area. The largest tangential stress of the SEI is still at the minor half-axis. In total, we demonstrate the importance of considering the mechanics of the anode and SEI in silicon anode simulations and encourage further numerical and model improvements.

Lithium metal batteries operated with high voltage cathodes are predestined for the realization of high energy storage systems, where solid polymer electrolytes offer a possibility to improve battery safety. Al2O3_PCL is introduced as promising hybrid electrolyte made from polycaprolactone (PCL) and Al2O3 nanoparticles that can be prepared in a one‐pot synthesis as a random mixture of linear PCL and PCL‐grafted Al2O3. Upon grafting, synergistic effects of mechanical stability and ionic conductivity are achieved. Due to the mechanical stability, manufacture of PCL‐based membranes with a thickness of 50 µm is feasible, yielding an ionic conductivity of 5·10⁻⁵ S cm⁻¹ at 60 °C. The membrane exhibits an impressive performance of Li deposition in symmetric Li||Li cells, operating for 1200 h at a constant and low overvoltage of 54 mV and a current density of 0.2 mA cm⁻². NMC622 | Al2O3_PCL | Li cells are cycled at rates of up to 1 C, achieving 140 cycles at >80% state of health. The straightforward synthesis and opportunity of upscaling as well as solvent‐free polymerization render the Al2O3_PCL hybrid material as rather safe, potentially sustainable and affordable alternative to conventional polymer‐based electrolytes.

In order to balance energy and power density of Li-ion batteries – and to ensure a long battery life – it is crucial to understand in detail how the composition and the microstructural design of the electrodes affects their capacity and kinetics. In our study we compare different commercial 21700 high-energy cells with very similar capacities (5 Ah) and energy densities (264 Wh/kg, 735 Wh/l) according to the supplier’s data sheets. However, in electrochemical tests, these cells showed significant differences with respect to rate capability and cyclic ageing. Detailed microstructural, chemical as well as electrochemical tests utilizing rebuilt half cells and symmetrical cells were done to identify microstructure-property relationships and the most critical microstructural parameters. In doing so, our study contributes to derive guidelines for optimized microstructural electrode design that combines energy with power and a long battery life. For material and microstructural analyses of fresh and aged cells, we used complementary methods such as optical and scanning electron microscopy, µCT- and FIB/SEM tomography, EDS and XRD to characterize the architecture of the electrodes at different length scales. We combined these results with results from detailed electrochemical studies such as OCV and half-cell tests with rebuilt cells as well as electrochemical impedance spectroscopy. Moreover, the 3D data from FIB/SEM tomography has been used as the input for microstructure-resolved simulations in the framework BEST, which allows for further investigation of the electrode microstructure and its effect on the electrochemical behaviour. In doing so, we gained a comprehensive dataset that allowed us to understand the different electrochemical behaviour of the different cells in the pristine state as well as their different behaviour during cyclic ageing. Some of the main findings are: (i) Different suppliers realize the same nominal capacity and energy densities of their cells by different electrode loadings; we observed differences of about 25 %; (ii) All suppliers strive to highly densify both, cathode and anode. Typical densities are 3.4 – 3.6 g/cm³ for the cathode and 1.5 – 1.6 g/cm³ for the anode. (iii) All investigated cells contain Si-rich phases within the anode, however, the chemical composition and the morphology is quite different and affects the electrochemical behaviour; (iv) We found significant differences in the binder-additive content; microstructure-resolved simulations indicate that this critically affects the electronic conductivity of the electrodes and thus, their electrochemical behaviour; (iv) The rate capability of the cells is significantly affected by the areal capacity of the electrodes (and presumably, their electronic conductivity); (vi) This in turn affects the cyclic aging behaviour of the cells. The investigated cells exhibit significant differences under various cycling protocols; we varied upper and lower cut-off voltage as well as the C-rate to investigate how microstructural features affect aging under different conditions. Within the talk, we will provide detailed insights into the results briefly outlined above and try to put the individual pieces of the puzzle together into an overall picture to elucidate how different microstructural features in combination determine the performance and cycling stability of the cells. Figure 1