High-fidelity hierarchical modeling of lithium-ion batteries: a cross-scale electrochemical-mechanical framework Authors: Xiaoyu Li, Shen Zhao, Chiton Gwan, Zhenpo Wang, Shuqiang Jiao, Yanli Zhu Published: 2025 ABSTRACT AbstractsMechanical stress during the cycling process notably impacts the performance of lithium-ion batteries (LIBs), making it crucial to accurately monitor stress generation and propagation during battery operation. Traditional electrochemical-mechanical models are limited to the particle and electrode scales, and their parameter identification relies solely on voltage. Here, a multi-scale electrochemical-mechanical-thermal modelling framework with non-destructive parameter identification capabilities is proposed. This numerical model couples electrochemical reactions with thermal effects and links particle-scale strain to electrode-scale displacement. Diffusion-induced stress (DIS) is selected as a key indicator, combined with voltage, to analyze the sensitivity of 23 parameters. A voltage-strain multi-objective parameter identification strategy based on the Pareto front is employed to determine the key parameters. The framework demonstrates high fidelity, with the mean absolute percentage error for voltage and strain predictions below 1% and 3.6%, respectively. This work enables high-fidelity simulation of multi-physics behavior, provides an effective method for calibrating key parameters, and holds potential for establishing a reliable digital twin of LIBs. Introduction Lithium-ion batteries (LIBs) have become the most crucial energy storage device in the current industry with wide applications from small portable devices to large-scale transportation and energy storage systems, mainly due to their high energy density, long cycle life and minimal self-discharge1–3. However, LIBs experience capacity decline during service due to degradation of the active material inside the battery and the occurrence of undesired side reactions. Among these factors, the phenomenon of contraction and expansion of the active particles of LIBs during charging and discharging, inducing multiscale mechanical deformations and stresses, is a major culprit in the capacity degradation and deterioration of the cycling performance of LIBs4–6. To this end, the understanding and modeling internal mechanical behavior of LIBs are essential and have been extensively studied. The effect of mechanical stress on the voltage hysteresis of LIBs during charging-discharging cycling has been demonstrated early on7. Diffusion-induced stresses (DIS), surface effects, and stresses from inter-particle compression all impede lithium embedding, resulting in energy dissipation and stress-dependent voltage hysteresis, thereby also affecting the electrochemical impedance of the battery8. Moreover, internal stress is an important contributor to capacity degradation. This is mainly manifested in two aspects. On one hand, during the charging and discharging process, the anode particles undergo significant volume changes due to the insertion and extraction of lithium ions (Li+). The solid electrolyte interface (SEI) formation breaks as the electrode particles expand, and the side reaction occurs again in the cracks, resulting in a continuous thickening of the SEI layer, reflecting as a continuous impedance increase and a capacity decrease9,10. On the other hand, Li+ insertion and extraction in the active particles form a lithium concentration gradient leading to DIS11,12, further contributing to the accumulation of elastic deformations in the anisotropic sphere called elastic strain energy density, providing a considerable driving force for the disruption of the active particles and the loss of active material13. Quantitative mapping between the electrochemical aging and mechanical stress is hard to be determined experimentally, which is further constrained by destructive testing methods and limitations in temporal resolution and spatial coverage, thus deviating from real-world application requirements14,15. Therefore, developing an electrochemical-mechanical coupling model is essential to accurately simulate the real behavior of the battery. Three kinds of models, namely electrochemical models (EM), equivalent circuit models (ECM) and fractional order models (FOM), are popular in battery simulation. ECM, as empirical models, employ resistor-capacitor-inductor networks to approximate electrochemical reactions16. FOM is a mathematical model based on fractional calculus theory, utilizing fractional derivatives to construct the kinetic characteristics of batteries. Compared to integer-order models, it offers superior accuracy in fitting nonlinear voltage responses and state-of-charge changes17,18. Similarly, FOM is also empirical, and its model parameters lack physical interpretability. The pseudo two-dimensional (P2D) model is one of the classical EM19. The high simulation accuracy, physical interpretability, and ability to seamlessly couple with other physical processes make it an ideal tool for studying electrochemical-mechanical behavior during lithium-ion battery operation, explaining its widespread application20,21. At the particle scale, 3-D electrochemical-mechanical models have been established to bridge the stress-strain relationship with lithium diffusion and transport processes22. Based on experimental data obtained from X-ray diffraction and swelling testers, a relationship between particle swelling and electrode thickness is established, laying the foundation for multi-scale mechanics23. Subsequent work further developed particle–electrode dual-scale mechanical models to analyze electrode mechanical mechanisms24 and safety application boundaries25. The mechanical framework was later extended to the full cell scale26,27, with thermal effects incorporated at this level to obtain one-dimensional temperature distributions28. Further refinement introduced porous electrode plasticity into a multi-scale one-dimensional thermal–mechanical model, enabling the calculation of cumulative electrode strain29. However, the internal electrochemical-mechanical parameters for these models are often derived from empirical values or simplified assumptions, potentially limiting their accuracy and predictive capability across diverse operating conditions. EM exhibits inherent nonlinearity. The P2D model and its derivatives contain many parameters with explicit physical meaning; obtaining the accurate values of these parameters is critical to ensure the accuracy of the model30. Therefore, a large number of studies have been carried out to identify the parameters for numerical simulations and experiments. Classified by the source of algorithmic inspiration, these parameter identification methods can be broadly classified into Mathematics-inspired, Evolutionary-inspired, Biology-inspired, Swarm-inspired and Human-inspired, etc. Building on this taxonomy, numerous algorithms have been developed for battery parameter identification. Mathematics-inspired and evolutionary-inspired approaches such as the genetic algorithm (GA) and ensemble Kalman filter have been used to identify 14 highly sensitive P2D parameters, achieving a voltage prediction RMSE below 14 mV31. The whale optimization algorithm (WOA), a swarm-inspired method, attained terminal voltage errors under 15.45 mV and SOC estimation errors below 1.21% when identifying 14 electrochemical parameters32. Other swarm-inspired algorithms including multi-objective particle swarm and grey wolf optimizers have accurately identified all 23 parameters of P2D models33. In a comprehensive benchmark study, Li et al.34 extensively counted and compared 78 common meta-heuristic algorithms for electrochemical parameter identification, and comprehensively evaluated the accuracy, convergence speed, and robustness of these algorithms, among which the teaching-learning-based optimization algorithm (TLBO), Multi-objective genetic algorithm (MultiGA), Bird Swarm Algorithm (BSA), Bald Eagle Search Algorithm (BES), Search and Rescue Optimization Algorithm and Cross-Entropy Method (CEM) algorithms perform best in terms of accuracy. It is worth noting that the GA, as one of the most classical optimization algorithms, still guarantees high accuracy in most application scenarios. In recent research, GA has been used in conjunction with machine learning to optimize the initial population and prevent it from trapping into local optima, with the ultimate goal of improving the accuracy of GA or speeding up the optimization process35,36. However, existing parameter identification frameworks for EM of LIBs primarily focus on electrochemical parameters, neglecting mechanical counterparts. Here, we propose a novel high-fidelity multiscale framework that integrates electrochemical-mechanical coupling model, sensitivity analysis and multi-objective optimization. Building upon the P2D model, the framework establishes quantitative linkages from particle-scale expansion to electrode-scale strain, and ultimately to cell-scale macroscopic displacement through multi-physics governing equations. This enables systematic characterization of DIS and its role in driving elastic strain energy density accumulation within active particles, thereby resolving mechanical-electrochemical coupling mechanisms across scales. Departing from conventional voltage-centric sensitivity analysis, hydrostatic stress is innovatively adopted as a key metric to evaluate parameter sensitivity, revealing critical mechanical parameters-including anode particle radius, Young’s modulus and electrode porosity-that dominate stress distribution and mechanical degradation. In parallel, a non-destructive multi-objective parameter identification strategy is developed by incorporating macroscopic strain measurements as an independent optimization target alongside terminal voltage. Leveraging Pareto front theory, four heuristic algorithms are comparatively evaluated. For multi-objective, MultiGA shows the fastest convergence speed and sufficiently high accuracy, with the fitness function converging to 0.0087, while the other three optimization algorithms converge to 0.0125, 0.0093, and 0.0086, respectively. The framework achieves mean absolute percentage errors (MAPE) below 1.0% for voltage and 2.9% for 1 C charging strain predictions, considerably enhancing the predictability of stress-induced capacity fade. Our method not only overcomes the limitations of destructive testing but also provides a validated tool for real-time battery management and electrode material design by correlating stress gradients with localized electrochemical activity. The framework and flowchart of the study are shown in Fig. 1.Fig. 1Schematic diagram of the multiscale coupling model parameter identification study.a Schematic diagram of multi-condition experiment (strain measurement method is shown on the left, and the results of the voltage strain section are shown on the right). b Schematic of multi-scale model (particle-to-cell scale model on the left, and partial simulation results on the right). c Sensitivity analysis of electrochemical-mechanical dual index (voltage and DIS response at different values of a parameter are on the left, radar diagram of sensitivity analysis results is on the right. DIS denotes diffusion-induced stresses). d Multi-objective parameter identification process through four heuristic optimization algorithms (GA stands for Genetic Algorithm, CEM represents Cross-Entropy Method, WOA is Whale Optimization Algorithm, and TLBO denotes Teaching-Learning-Based Optimization algorithm). Schematic diagram of the multiscale coupling model parameter identification study. a Schematic diagram of multi-condition experiment (strain measurement method is shown on the left, and the results of the voltage strain section are shown on the right). b Schematic of multi-scale model (particle-to-cell scale model on the left, and partial simulation results on the right). c Sensitivity analysis of electrochemical-mechanical dual index (voltage and DIS response at different values of a parameter are on the left, radar diagram of sensitivity analysis results is on the right. DIS denotes diffusion-induced stresses). d Multi-objective parameter identification process through four heuristic optimization algorithms (GA stands for Genetic Algorithm, CEM represents Cross-Entropy Method, WOA is Whale Optimization Algorithm, and TLBO denotes Teaching-Learning-Based Optimization algorithm). Results and discussion Experimental platform and results In this study, commercial small pouch LIBs are used for the test, with graphite anodes and LiNi0.5Co0.2Mn0.3O2 (NCM523) cathodes. The capacity of the battery is 4 Ah, the nominal voltage is 3.7 V, and the dimensions are 79 mm × 40 mm × 10 mm. The experiments were conducted under three distinct operating conditions: 1 C charging, 0.05 C discharging, and the Urban Dynamometer Driving Schedule (UDDS). These conditions were selected to isolate kinetic-dominated and diffusion-dominated behaviors, as well as to simulate real-world stochastic driving patterns, ensuring robust parameter identifiability and enhancing the generalizability of the cross-scale model. Experimental equipment can be found in the Supplementary Information. (Supplementary Fig. 9). The experimental setup included a high-precision battery cycler manufactured by Neware (model CT-8002), a voltage sensor for real-time voltage monitoring, and a resistance strain gauge made of Constantan with a resistance of 120 Ω. The strain gauge, featuring a micron-sized foil structure, was placed at the center of the battery surface to measure surface strain caused by volume changes during cycling. A fixture was used to clamp the battery horizontally, applying an initial external pressure of 100 N to eliminate residual stresses and ensure accurate stress-strain transfer. The strain gauge’s design, characterized by low mass-volume and non-invasive embedding, minimized additional load on the battery, making it suitable for miniaturized and flexible structures like LIBs electrodes. Its high linearity, wide frequency response, and advanced temperature compensation algorithms enabled precise measurement of micro-strain deformations, effectively suppressing ambient noise. This capability allowed for in situ strain monitoring at the electrode-electrolyte interface and internal stress field reconstruction under packaging constraints, making it an ideal tool for studying mechanical-electrochemical coupling behavior. Multiscale electrochemical-mechanical coupling behavior The established electrochemical-mechanical coupling model offers a comprehensive understanding of the multi-physical behaviors of LIBs, greatly enhancing the accuracy of performance prediction. This mechanism-based framework not only captures the external characteristics of the battery, such as voltage profiles, but resolves internal phenomena, including concentration gradients and polarization effects. Crucially, it quantifies the volumetric strain induced by lithium concentration gradients and the resulting stress distribution, enabling a deeper insight into stress-driven degradation mechanisms. The current input from the pole lugs is transferred to the surface of the electrode active particles through the collector, triggering an electrochemical reaction at the electrode/electrolyte interface. Li+ is subsequently inserted/extracted within the particles via solid-phase diffusion, thereby establishing a concentration gradient between the particle surface and core. Insertion/extraction of solute atoms changes the lattice constant. Insertion of Li+ in graphite leads to expansion of the interlayer spacing. The diffusion of Li+ is driven by a concentration gradient, resulting in local volume changes in the material. The lithium concentration on the surface of the electrode material may be higher than in the interior, causing the surface to expand more and the interior to expand less, this uneven expansion creates stresses within the particle. Fig. 2 illustrates the electrochemical-mechanical behavior of the particles during 1 C charging, and the behavior of the particles for the other two conditions are presented in Supplementary Information (Supplementary Fig. 2).Fig. 2Particle-scale electrochemical-mechanical behavior during 1 C charging.a Anode particle radius (left y-axis), cathode particle radius (right y-axis). b Average embedded particle concentration at anode and cathode. c Hydrostatic stresses within the particles at a location on the anode, black side for early charging, copper side for final charging. d Variation of hydrostatic stress inside anode particles with the concentration of embedded particles. e Variation of total elastic strain energy density of anode particles with time (polar coordinate system). f Distribution of total elastic strain energy density of anode particles at the electrode (Cartesian coordinate system). Particle-scale electrochemical-mechanical behavior during 1 C charging. a Anode particle radius (left y-axis), cathode particle radius (right y-axis). b Average embedded particle concentration at anode and cathode. c Hydrostatic stresses within the particles at a location on the anode, black side for early charging, copper side for final charging. d Variation of hydrostatic stress inside anode particles with the concentration of embedded particles. e Variation of total elastic strain energy density of anode particles with time (polar coordinate system). f Distribution of total elastic strain energy density of anode particles at the electrode (Cartesian coordinate system). The initial radius of the identified graphite anode particles is larger than that of the NCM cathode particles, and the radius change of the anode particles is much larger than that of the cathode particles during the charging and discharging process, as shown in Fig. 2a. The change in the radius of the particles is basically the same for a single complete discharging or charging up to the cut-off voltage, even though the size of the input currents varies greatly. The variation in embedded particle concentration is highly consistent with that in particle radius under the three conditions, as clearly seen in the UDDS dynamic discharging condition. The results of particle-scale electrochemical-mechanical behavior can be found in Supplementary Information (Supplementary Fig. 2), which demonstrates that the radius change is driven by concentration, also shown in Fig. 2b. The radial DIS reaches its maximum compressive stress at the center of the particle during early charging, driven by the steep Li+ concentration gradient forming near the surface while the center remains less lithiated. This central compressive stress increases rapidly to a peak value of 51 MPa as charging progresses. Subsequently, as Li+ diffuses inward, reducing the concentration gradient, the central compressive stress begins to diminish. Near the end of the constant current charging phase, the central radial stress slowly rises and then begins to show a decreasing trend as lithiation progresses towards homogeneity. The tangential DIS distribution evolves significantly during charging. Initially, high compressive stress develops at the center, reaching about 50 MPa, while simultaneously, constrained surface expansion generates high tensile stress at the particle periphery. As charging continues, the radial position where tangential stress is zero gradually migrates inward toward the center. Concurrently, the magnitude of the surface tensile stress decreases from its peak. If the charging continues, as lithium concentration becomes more uniform, the surface tensile stress further decreases and the overall tangential stress distribution becomes more homogeneous. This mechanism of DIS, governed by evolving lithium concentration gradients and associated deformation constraints, is further validated under discharge conditions. During both 0.05 C and UDDS discharge tests achieving nearly full discharge, the absolute values of radial and tangential DIS consistently exhibited an initial increase followed by a subsequent decrease. While the specific stress states are reversed during delithiation compared to charging, the pattern of stress evolution, with an initial rise driven by developing concentration gradients and surface deformation constraints, followed by relaxation as gradients diminish. While the specific stress states are reversed during delithiation compared to charging, the fundamental stress evolution pattern remains consistent, featuring an initial increase driven by developing concentration gradients and surface deformation constraints, followed by stress relaxation as these gradients diminish. (All conditions radial and tangential DIS are available in Supplementary Fig. 1). Figure 2c, d show the hydrostatic stress and its relationship with the embedded particle concentration during 1 C charging. The hydrostatic stress is a weighted sum of radial DIS and tangential DIS, which is mainly dominated by the tangential direction, thus, the stress distribution is similar to that of the tangential direction, with maximum stresses of 59 MPa at the center side and −33 MPa at the edge side of the particles, respectively. The elastic strain energy density is introduced to describe the cumulative elastic deformation of the spheres, also known as the strain energy per unit volume. The total elastic strain energy density \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${W}_{s,{{\rm{tot}}}}$$\end{document}Ws,tot stored in the electrode material provides the driving force for the rupture of the particle surface, which can be used as an indicator to solve for the loss of active material in subsequent long-term charging/discharging cycling studies, as shown in Fig. 2e, f. Overall, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${W}_{s,{{\rm{tot}}}}$$\end{document}Ws,tot increases as charging proceeds, with a maximum \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${W}_{s,{{\rm{tot}}}}$$\end{document}Ws,tot of 1.78 × 106 J·m−3. The difference in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${W}_{s,{{\rm{tot}}}}$$\end{document}Ws,tot between different particles on the negative electrode reaches a maximum at 1400 s, with a maximum difference of 2.66 × 105 J·m−3. This inhomogeneous strain energy is mainly driven by the concentration gradient and the inhomogeneous current distribution. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${W}_{s,{{\rm{tot}}}}$$\end{document}Ws,tot is uniformly distributed on the negative electrode. The edge side of the particles is gradually higher than the inside of the cell as charging proceeds, with a maximum difference of 0.24 MPa. Particles close to the collector preferentially bear higher current densities due to short electron transport paths, leading to faster Li+ insertion/extraction rates and notably higher local reactivity than particles in the middle of the electrode. Similarly, Li+ needs to migrate from the electrolyte across the separator to the opposite electrode, the ion concentration gradient of the electrolyte near the separator is the largest, in addition, the electrolyte near the separator is susceptible to lithium depletion (when discharging) or lithium enrichment (when charging), which leads to an increase in the concentration overpotential and drives a higher local reactive current. Fig. 3a shows the local current density of the negative electrode.Fig. 3Electrode scale electrochemical-mechanical typical behavior during 1 C charging.a Local current density distribution at the negative electrode. b overpotential distribution on the electrode. c relationship between overpotential and local current density (The sub-insets show local magnified views of overpotential in the ranges of −2.2 to 1.1 A·m−2 and 0.3 to 0.55 A·m−2.). d electrolyte current density on the electrode (Elyte is an abbreviation for electrolyte. The sub-inset shows local magnified views of the electrolyte current density in the range of 3.8 to 4 × 10−5 m on the electrode.). e V-M stress at the particle surface-electrode interface on the negative electrode (V-M denotes von Mises). f V-M stress inside the particle somewhere on the negative electrode. Electrode scale electrochemical-mechanical typical behavior during 1 C charging. a Local current density distribution at the negative electrode. b overpotential distribution on the electrode. c relationship between overpotential and local current density (The sub-insets show local magnified views of overpotential in the ranges of −2.2 to 1.1 A·m−2 and 0.3 to 0.55 A·m−2.). d electrolyte current density on the electrode (Elyte is an abbreviation for electrolyte. The sub-inset shows local magnified views of the electrolyte current density in the range of 3.8 to 4 × 10−5 m on the electrode.). e V-M stress at the particle surface-electrode interface on the negative electrode (V-M denotes von Mises). f V-M stress inside the particle somewhere on the negative electrode. The local current density is regulated by overpotential. The nature of overpotential is the extra energy required to overcome the reaction energy barriers. In the charging and discharging process, overpotentials are inevitably generated due to the kinetic limitations of the electrochemical reaction, the resistance of the material and the electrolyte and the concentration gradient during the migration of Li+. In the macroscopic electrochemical characterization, overpotentials are modulated by the solid/liquid phase potential difference and the open circuit voltage, as well as affected by hydrostatic stresses. The overpotentials and the relationship with the local current density during 1 C charging are shown in Fig. 3b, c. This plot visually demonstrates the response characteristics of local current density to overpotential variation, providing profound insights into the inherent kinetics of charge transfer reactions at the electrode/electrolyte interface and their underlying limiting mechanisms. Notably, during constant-current charging, the stable Tafel slope observed in the linear regime reflects consistent charge-transfer kinetics. The electrolyte current density, determined by ionic conductivity and concentration gradients, ensures charge conservation by matching the local current density at the interface. The distribution of electrolyte current density on the electrode is shown in Fig. 3d. The spatial heterogeneity of the local current density during charging leads to inhomogeneous lithium intercalation, which generates large concentration gradients within the active particles. These gradients induce chemical strains, which in turn lead to the accumulation of surface stresses. The current density peaks near the separator-electrode interface due to the increased ion transport efficiency. This localized electrochemical activity interacts with elevated V-M stresses in the same region. The mechanical behavior of materials is more complicated to analyze directly employing the stress tensor, especially under multi-axial stress states. Von Mises (V-M) stress, as a scalarized equivalent stress, is able to comprehensively reflect the deformation energy of materials under complex stress states, highlighting critical regions prone to failure such as particle surfaces and electrode interfaces. Figure 3e demonstrates the V-M stress distribution on particle surfaces across the negative electrode. Throughout the initial 1770 s of charging, V-M stress exhibited an increasing trend. The stress near the separator side rose markedly faster than at the current collector side, reaching a peak value of 53 MPa. This accelerated stress accumulation stems from enhanced ion transport efficiency near the separator, which elevates local current density, accelerates Li+ intercalation, and generates steeper concentration gradients that induce severe chemical strain. After 1770 s, as active particles approached Li+ saturation near the separator and concentration gradients relaxed, V-M stress decreased abruptly. In contrast, stress at the collector side continued rising to 33 MPa due to delayed reaction kinetics and mechanical constraint from the current collector. Charging ended at this point; extended charging would likely drive the collector-side stress to peak before declining, ultimately stabilizing at a low plateau upon homogenization of concentration gradients. Figure 3f illustrates the V-M stress distribution within a particle. Stress remained zero at the core but increased rapidly near the edge where lithium intercalation initiates, creating maximal concentration gradients and heterogeneous expansion. As charging progressed, Li+ diffusion toward the particle core dynamically redistributed strain, elevating overall stress. Beyond 1950s, the inward migration of the concentration gradient peak reduced the maximum stress growth rate, culminating in a declining trend. Electrode-scale electrochemical-mechanical behavior under 0.05 C and UDDS conditions can be found in Supplementary Information (Supplementary Fig. 3). From Eqs. (39) to (41) in Supplementary Information (Supplementary Equation 2), the electrode-scale mechanics are extended to the cell scale. The simplification assumes that the concentration size and distribution are the same in all cells, with neighboring cells sharing a common collector, for a total of 57 cells. Consider that one side of the cell is fully fixed and the other side is preloaded with a preload force of 100 N applied through a fixture. The further away from the fixed side of the cell, the displacement gradually accumulates, but the pressure does not accumulate considerably, or the pressure generated by the negative electrode is much larger than the accumulated amount. The view is magnified by a factor of 20 in the thickness direction. The displacement of the cell is dominated by the expansion of the negative electrode, with a maximum displacement of 9.5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mu m$$\end{document}μm inside the cell and 6.4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mu m$$\end{document}μm at the boundary. The pressure generated by the expansion of the negative electrode region is much higher than that of the other regions, with a maximum pressure of 818 MPa inside and 78 MPa at the boundary. Cell-scale displacements, strain and stresses are displayed in Fig. 4.Fig. 4Cell-scale mechanical characterization.a Boundary displacement curve and cloud diagram. b Boundary pressure curve and cloud diagram. c Comparison of strain between the purely elastic and elastic-plastic models under 1 C charging conditions, along with the equivalent plastic strain (right y-axis). d Comparison of strain between the purely elastic and elastic-plastic models under 0.5 C charging conditions, along with the equivalent plastic strain. Cell-scale mechanical characterization. a Boundary displacement curve and cloud diagram. b Boundary pressure curve and cloud diagram. c Comparison of strain between the purely elastic and elastic-plastic models under 1 C charging conditions, along with the equivalent plastic strain (right y-axis). d Comparison of strain between the purely elastic and elastic-plastic models under 0.5 C charging conditions, along with the equivalent plastic strain. The purely elastic model can accurately simulate low-rate charge-discharge tests over relatively short durations. However, during long-term cycling or high-rate tests, the electrodes undergo marked plastic deformation, which is a key focus of our subsequent efforts. Here, a relatively homogeneous equivalent yield stress is introduced to represent the overall plasticity of the electrode, and this equivalent yield stress value is obtained through multi-objective parameter identification, highlighting one of the advantages of parameter identification. Under both 1 C and 0.5 C charging conditions, equivalent plastic strain first appears when the half-cell SOC is around 0.55 (Fig. 4a, b). The elastoplastic-viscous model predicts 3.9% and 3.5% larger electrode swelling than the purely elastic model under 1 C and 0.5 C conditions, respectively, attributable to lithiation-induced stress exceeding the yield limit causes irreversible plastic deformation in graphite. Additionally, viscoelastic effects promote plastic flow through stress relaxation, and the two effects synergistically exacerbate strain accumulation. During the charging and discharging process of LIBs, the insertion/extraction behavior of Li+ triggers the entropy effect through the change of lattice structure of the active material, which generates reversible heat, with its exothermic or absorptive properties determined by the direction of the reaction. Meanwhile, the current flow through the electrode and electrolyte is subject to the effect of ohmic impedance and interfacial kinetic resistance, and irreversible heat is formed through the dissipation of joule heat and polarization energy. The reversible heat is directly related to the intrinsic thermodynamic properties of the material, while the irreversible heat originates from the multiple loss mechanisms in the energy transfer process, and the synergistic effect of the two together constitutes the basis of the thermal behavior of the battery during operation. The distribution of heat production of LIBs reaction during 1 C charging are provided in Supplementary Information (Supplementary Fig. 4). The temperature changes and temperature distribution under the 1 C charging condition are shown in Fig. 5, and the temperature behaviors of the other two conditions are given in Supplementary Information (Supplementary Fig. 5).Fig. 5Cell-scale thermal characterization.a Temperature changes with charging progress. b Surface temperature distribution at the end of charging. Cell-scale thermal characterization. a Temperature changes with charging progress. b Surface temperature distribution at the end of charging. Sensitivity analysis results Selected 20 conventional electrochemical parameters cover all electrochemical processes of the P2D model. Each physical parameter individually affects the model voltage output, i.e., the terminal voltage at the designed current input, which indicates the sensitivity of the parameter. However, these parameters are not necessarily sensitive to the mechanical behavior during charging and discharging, and more parameters are introduced when considering the mechanical behavior, ignoring material damage and plasticity for short-term operating conditions, the main mechanical parameters include Young’s modulus, Poisson’s ratio and partial molar volume of the particles. In this way the parameters of the whole electrochemical-mechanical process are considered comprehensively. The sensitivity analysis radar diagrams of 23 parameters are shown in Fig. 6. When hydrostatic stress is used as an indicator, three parameters that have almost no effect in all three conditions are deleted and replaced by three mechanical parameters of the particles; the replaced parameters are\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon }_{e}^{s}$$\end{document}εes, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${D}_{e}$$\end{document}De and\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${c}_{e}$$\end{document}ce, respectively.Fig. 6Electrochemical-mechanical parameter sensitivity analyses.a Terminal voltage as an indicator, the right subplot is magnified to 0-0.11 (UUDS stands for Urban Dynamometer Driving Schedule). b Hydrostatic stress as an indicator, the right subplot is magnified to 0-0.11. Electrochemical-mechanical parameter sensitivity analyses. a Terminal voltage as an indicator, the right subplot is magnified to 0-0.11 (UUDS stands for Urban Dynamometer Driving Schedule). b Hydrostatic stress as an indicator, the right subplot is magnified to 0-0.11. When terminal voltage serves as the sensitivity metric, the volume fraction and porosity of the composite electrode exhibit the highest dominance under 1 C charging and UDDS conditions, followed by the geometric dimensions of the positive and negative electrodes. In contrast, the 0.05 C discharging condition prioritizes the negative electrode particle radius as the decisive parameter, with the negative electrode diffusion coefficient ranking second in sensitivity. The parameter dominance hierarchy becomes more pronounced when hydrostatic stress functions as the evaluation indicator. The anode particle radius and Young’s modulus play a decisive role in the magnitude of stress generation. The highest normalized sensitivity coefficients of the anode particles are found in the 1 C charging and 0.05 C discharging conditions. For UDDS conditions, Young’s modulus demonstrates the highest normalized sensitivity coefficient, followed by the negative electrode particle radius, with a value of 0.482. Notably, the negative electrode diffusion coefficient retains its dominance in the 0.05 C discharging scenario, underscoring its pivotal role in low-rate diffusion-governed regimes. Additionally, the equivalent yield stress of the electrode is also identified. Since yield stress belongs to electrode-scale mechanics, while DIS reflects particle-scale mechanics, yield stress fails to be considered in sensitivity analysis. In subsequent studies, an indicator sensitive to both multi-scale electrochemical and mechanical behavior will be explored. Parameter identification results Four heuristic optimization algorithms are employed to perform dual objective parameter identification of the 1 C charging results combining voltage and strain, the comparison of the fitness curves is shown in Fig. 7a. Due to the limitation of the test conditions, it is not possible to accurately measure the strain for 0.05 C discharging and dynamic conditions, so only the voltage single objective is used. Since the mechanical parameters are few and sensitive to the results of 1 C charging, accurate parametric results can be recognized from a single condition only. The WOA, which has the highest accuracy in 1 C, is used for parameter identification for the 0.05 C discharging and UDDS conditions, and the fitness curves are shown in Fig. 7b.Fig. 7Fitness curves for three conditions.a Comparison of fitness curves of four algorithms for 1 C charging condition (CEM represents Crossed Entropy Method, TLBO represents Teaching-Learning-Based Optimization, WOA represents Whale Optimization Algorithm, and MultiGA represents Multi-Objective Genetic Algorithm.). b Fitness curves for 0.05 C and UDDS conditions (UUDS represents Urban Dynamometer Driving Schedule.). Fitness curves for three conditions. a Comparison of fitness curves of four algorithms for 1 C charging condition (CEM represents Crossed Entropy Method, TLBO represents Teaching-Learning-Based Optimization, WOA represents Whale Optimization Algorithm, and MultiGA represents Multi-Objective Genetic Algorithm.). b Fitness curves for 0.05 C and UDDS conditions (UUDS represents Urban Dynamometer Driving Schedule.). MultiGA explores extensively in the solution space through crossover and mutation operations, with a high initial population diversity that enables rapid discovery of multiple potential non-dominated solutions, using non-dominated ordering and crowding distances to preserve the diversity of the Pareto front and avoid premature convergence. In the early iterations, the set of non-dominated solutions expands rapidly and the fitness decreases markedly. Thus has the fastest convergence rate. the teacher phase of TLBO may repeatedly pull the population towards the local frontiers of different objectives, while the stochastic interactions in the student phase cannot effectively coordinate the multi-objective trade-offs, leading to fluctuations in fitness. the mean of the CEM may be biased towards the optimal region of Objective 1 in one iteration, and then be shifted in the next iteration due to the predominance of elite samples in Objective 2, leading to fitness oscillations. In complex multi-peak problems, WOA can adaptively adjust the search strategy to avoid falling into the local optimum prematurely, while guaranteeing high-precision convergence at a later stage, and WOA has the fastest computational efficiency. Therefore, WOA is used for parameter identification for the 0.05 C and UDDS conditions. Obviously, the difference between experimental and simulated of CEM identification results is higher than the other two algorithms, with the maximum absolute error of voltage being 65.8 mV and the maximum absolute error of strain being −2.04 × 10−5. The maximum error of voltage for the remaining two algorithms is 36.4 mV, 42.2 mV, and the maximum error of strain is 1.54 × 10−5 and 9.85 × 10−6, respectively. Combining the results of the dual-objective, intuitively, MultiGA is more accurate than TLBO, both of which are better than CEM in terms of accuracy and convergence speed. This may be due to CEM relies on the elite samples to update the distribution parameters, but in the dual-objective problem the elite samples may be concentrated in the vicinity of the locally optimal solution of a certain objective, ignoring the direction of optimization of the other objective, also, it is prone to fall into local optima in nonlinear and nonconvex problems. For 1 C charging, the absolute error is less than 21.3 mV at all points except the initial value, and for 0.05 C discharging, the absolute error is less than 41.5 mV at all moments. For UDDS conditions, the absolute error is periodic as the current varies periodically. When the amplitude variation is slight, as shown in the subplot of Fig. 8i, it is able to maintain the difference within 20 mV of the absolute value, but when the amplitude is large the voltage response of the simulation is slightly smaller than that of the experiment. This discrepancy likely originates from intensified polarization under dynamic current conditions, where internal parameter instability leads to inherent limitations in fixed-value simulations.Fig. 8Comparison of experimental and simulation results.a 1 C charging voltage employing CEM (CEM represents Crossed Entropy Method). b 1 C charging strain employing CEM. c 1 C charging voltage employing TLBO (TLBO stands for Teaching-Learning-Based Optimization). d 1 C charging strain employing TLBO. e 1 C charging voltage employing MultiGA (MultiGA refers to Multi-Objective Genetic Algorithm). f 1 C charging strain employing MultiGA. g 1 C charging voltage employing WOA (WOA denotes Whale Optimization Algorithm). h 1 C charging strain employing WOA. i 0.05 C discharging voltage employing WOA. j UDDS condition voltage employing WOA. (The error on the right y-axis represents absolute error). The corresponding specific data are available in Supplementary Data 1. Comparison of experimental and simulation results. a 1 C charging voltage employing CEM (CEM represents Crossed Entropy Method). b 1 C charging strain employing CEM. c 1 C charging voltage employing TLBO (TLBO stands for Teaching-Learning-Based Optimization). d 1 C charging strain employing TLBO. e 1 C charging voltage employing MultiGA (MultiGA refers to Multi-Objective Genetic Algorithm). f 1 C charging strain employing MultiGA. g 1 C charging voltage employing WOA (WOA denotes Whale Optimization Algorithm). h 1 C charging strain employing WOA. i 0.05 C discharging voltage employing WOA. j UDDS condition voltage employing WOA. (The error on the right y-axis represents absolute error). The corresponding specific data are available in Supplementary Data 1. The violin plots of the relative errors of the voltages recognized by the three algorithms are shown in Fig. 9a. This composite figure combines a kernel density map (colored shaded region representing data distribution) with a box plot (white square center and dashed lines at upper/lower ends, where dashed regions denote outliers). The absolute relative error for all points is less than 2.1%. The CEM error result exists a peak at 0.823%, the median is marked by a red horizontal line, and the median relative error is 0.603%. In contrast, the TLBO error result exists two peaks of similar density with similar densities at 0.03% and 0.29%, respectively, with a median of 0.010% and the upper and lower limits of the boxplot ranging from −0.87% to 0.68%. The MultiGA error result exists one distinct peak positioned at −0.19% with a median of −0.19% and the boxplot ranging from −0.66% to −0.17%, indicating a more concentrated distribution. All three algorithms achieve MAPE below 0.3% for terminal voltage predictions. Particularly, TLBO and MultiGA approach near-optimal performance with MAPE values of 0.22% and 0.21%, respectively. Corresponding RMSE remains low at 0.0108 V and 0.0111 V for these two algorithms.Fig. 9Error analysis of terminal voltage.a Voltage relative error employing different algorithms. b Voltage MAPE and RMSE employing different algorithms. c Voltage relative error in three conditions employing WOA. d Voltage MAPE and RMSE in three conditions employing WOA. Error analysis of terminal voltage. a Voltage relative error employing different algorithms. b Voltage MAPE and RMSE employing different algorithms. c Voltage relative error in three conditions employing WOA. d Voltage MAPE and RMSE in three conditions employing WOA. The violin plots of the relative voltage errors for the three conditions are shown in Fig. 9c. The width of the violin plots reflects the kernel density estimates of the data distributions, with peaks occurring within 0.5% for all three conditions. The upper and lower limits of the box plot for the charging condition are 0.098% and −0.91%, respectively. The violin plots for the 0.05 C discharging condition show the presence of two more pronounced peaks, with a range of box plots of −1.07% to 1.22%. The median (red line) of the UDDS condition box plot is -0.15% with a range of −1.75% to 1.57%. The optimized parameters are presented in the supplementary information (Supplementary Table 1). The optimal identification results are the initial values of the parameters, some of which will be changed as charging and discharging proceeds, controlled by the driving equations in Section III. As can be seen, none of the parameters take values up to the boundary values of the range, which proves the credibility and reasonableness of the identification results. To validate the robustness of the model in terms of mechanical properties, 0.5 C constant-current charging tests are supplemented to capture the voltage and strain responses under milder kinetic conditions. This multi-rate validation strategy aims to assess whether there is a notably current dependence of the electrochemically coupled mechanical parameters. While simultaneous optimization for both 0.5 C and 1 C datasets is ideal to ensure broad applicability, its computational cost is prohibitive for the current high-fidelity multiphysics model, as well as the inconsistent concerns of parameter sensitivity. The electrochemical process and the mechanical response are strongly coupled, and the particle- and electrode-scale mechanical parameters can be satisfied accordingly while ensuring accurate voltage estimation for the three representative working conditions. Compared to the electrochemical parameters, the mechanical parameters are fewer, and the relationship from expansion to displacement can be obtained with a single determination from a single working condition. The cross-scale mechanical relationship and the electrode material mechanical characteristics are unable to be explored by the parameter identification method, which relies on more detailed experiments and more elaborate mechanism exploration, which is also the direction of our future work. A comparison of the simulation and test results for 0.5 C is shown in Supplementary Information (Supplementary Fig. 7). The model jointly identified using 1 C charging, 0.05 C discharging, and UDDS conditions well reproduces the key features of 0.5 C charging. 0.5 C charging simulation voltage shows a maximum absolute error of 19.4 mV, and the RMSE is 0.0121 V, which is only 6.1% larger compared with 0.0114 for 1 C charging. The experiment excludes the constant-voltage charging stage, since higher rate implies larger overpotential. In contrast, the actual potential of the half-cell during 0.5 C charging up to the cutoff voltage is closer to its open-circuit voltage, and the actual lithium embedded in the material is larger, thus the final surface strain is also slightly larger than that of the 1 C charging. Methods Model geometry and framework As an effective electrochemical simulation tool, the P2D model is widely used to analyze the performance of electrode materials and the overall battery behavior. The P2D model can effectively describe Li+ intercalation and diffusion behavior inside the electrode, as well as the electrochemical reaction mechanisms. It focuses particularly on the interactions between the particle intercalation and the porous electrodes, which is crucial for understanding the battery charging and discharging process and optimizing its performances. Moreover, the equation structure of the P2D model can tightly couple the electrochemical, thermal and mechanical mechanisms through explicit source terms and intrinsic relationships, while retaining the independence, which facilitates numerical solution and analysis of physical mechanisms. All governing equations in this study are implemented in finite element simulations to determine the electrochemical and mechanical response of the battery. The mathematical equations are solved numerically using COMSOL 6.1. The model geometry and the multi-scale framework are shown schematically in Supplementary Information (Supplementary Fig. 8). Particle to cell scale equations The multi-scale electrochemical-mechanical coupling process of LIBs is a complex multi-physics problem involving electrochemical reactions, Li+ transport, electron conduction and stress-strain distribution. This process spans from microscopic particles to macroscopic electrodes and even the full cell, exhibiting markedly multi-scale characteristics and strong coupling effects. The electrochemical process primarily includes the migration and diffusion of Li+ in the electrolyte, the diffusion of Li+ in solid active particles, charge conduction and the electrochemical reactions occurring on the surface of particles. Furthermore, changes in lithium-ion concentration induce volumetric expansion and contraction of particles, leading to complex stress-strain fields, which in turn influence the electrochemical reaction kinetics and lithium diffusion processes. The fidelity of simulation results is critically dependent on the granularity of the model. Referring to previous excellent electrochemical-mechanical related studies27,37–39, the multi-scale coupled physical model is developed. The coupled model integrates Li+ diffusion, electrochemical kinetics and stress-strain interactions across particle-to-cell scales. By bridging microscopic particle deformation with macroscopic electrode displacement, the framework resolves bidirectional stress-electrochemical coupling mechanisms. Key governing equations encompass Li+ diffusion in solid particles, stress-strain constitutive relationships and electrode-scale mechanical balance, enabling quantitative analysis of DIS and its role in capacity fade. Detailed derivations are provided in Supplementary Information (Supplementary Equation 1). This hierarchical approach ensures high-fidelity predictions of voltage-strain behavior under dynamic operating conditions. The diffusion of lithium ions in active particles is described by Eq. (1).1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{\partial {c}_{s}}{\partial t}+\nabla \cdot j=0$$\end{document}∂cs∂t+∇⋅j=0where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${c}_{s}$$\end{document}cs represents the solid Li+ concentration in the electrode particles, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$j$$\end{document}j represents the flux passing through the solid-liquid interface, characterizes the mass transfer rate. Spherical particles are considered as isotropic linear elastic materials. When considering the stress caused by Li+ diffusion, the stress-strain constitutive equation of the particles is determined by Eq. (2)39, which incorporates the effects of concentration gradients and elastic deformation.2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon }_{ij}=\frac{1}{E}[(1+\nu ){\sigma }_{ij}-\nu {\sigma }_{kk}{\delta }_{ij}]+\frac{\Delta {c}_{s}\varOmega }{3}{\delta }_{ij}$$\end{document}εij=1E[(1+ν)σij−νσkkδij]+ΔcsΩ3δijwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon }_{ij}$$\end{document}εij are strain components, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{ij}$$\end{document}σij are stress components, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E$$\end{document}E is Young’s modulus of the active particle, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\nu$$\end{document}ν is Poisson’s ratio of the active particle, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\delta }_{ij}$$\end{document}δij is Kronecker delta. Since the mechanical balance is a fast process relative to Li+ diffusion, the electrode particles can be regarded as a quasi-static mechanical balance system at any time, and the stress balance equation is given by Eq. (3).3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{{\rm{d}}}{\sigma }_{{{\rm{r}}}}}{\,{{\rm{d}}}r}+\frac{2}{r}({\sigma }_{{{\rm{r}}}}-{\sigma }_{t})=0$$\end{document}dσrdr+2r(σr−σt)=0where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r$$\end{document}r is the particle radius, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{r}$$\end{document}σr represents radial stress and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{t}$$\end{document}σt denotes tangential stress. Electrodes experience considerably expansion during Li+ insertion/extraction, particularly graphite anodes, which inevitably lead to plastic flow during long-term cycling. Additionally, the viscoelastic relaxation effects at the particle-binder interface also influence the stress evolution patterns. These two non-elastic deformation mechanisms collectively determine the mechanical response and cycling stability of the electrode. Based on the continuum mechanics framework, the deformation gradient of electrode materials can be decomposed into elastic, plastic, chemical expansion, and thermal expansion components, which is given by Eq. (4).4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\boldsymbol{F}}}{=}{{{\boldsymbol{F}}}}_{e}{{{\boldsymbol{F}}}}_{p}{{{\boldsymbol{F}}}}_{ch}{{{\boldsymbol{F}}}}_{th}$$\end{document}F=FeFpFchFthwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\boldsymbol{F}}}}_{e}$$\end{document}Fe represents the deformation gradient caused by elastic deformation of the electrode, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\boldsymbol{F}}}}_{p}$$\end{document}Fp denotes the deformation gradient of the plasticity of the electrode, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\boldsymbol{F}}}}_{ch}$$\end{document}Fch denotes the lithiation-induced deformation gradient and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\boldsymbol{F}}}}_{th}$$\end{document}Fth represents the deformation gradient induced by temperature change. Specific electrode mechanical details are shown in Supplementary Information (Supplementary Equation 2)40,41. The electric double-layer capacitance, which is the capacitance effect due to charge separation at the electrode-electrolyte interface, describes the effect of the electric double-layer on the electrochemical kinetics at the electrode surface. The governing equation for the current in the solid phase substrate when considering the electric double-layer capacitance is given by Eq. (5).5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\nabla {i}_{{{\rm{s}}}}=-{a}_{v}\left(Fj+{C}_{{{\rm{dl}}}}\left(\frac{\partial {\phi }_{{{\rm{s}}}}}{\partial t}-\frac{\partial {\phi }_{{{\rm{l}}}}}{\partial t}\right)\right)$$\end{document}∇is=−avFj+Cdl∂ϕs∂t−∂ϕl∂twhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${a}_{v}$$\end{document}av represents specific surface area of active particles, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{{dl}}$$\end{document}Cdl indicates the electric double-layer capacitance, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\phi }_{{{\rm{s}}}}$$\end{document}ϕs is solid-phase potential and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\phi }_{\,{{\rm{l}}}}$$\end{document}ϕl is electrolyte-phase potential. The electrochemical reaction takes place at the interface between electrolyte and electrode, which is controlled by Bulter-Volmer equation as Eq. (7).6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$J={a}_{v}{i}_{0}\left[\exp \left(\frac{{\alpha }_{a}F}{RT}\eta \right)-\exp \left(-\frac{{\alpha }_{c}F}{RT}\eta \right)\right]$$\end{document}J=avi0expαaFRTη−exp−αcFRTηwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${i}_{0}$$\end{document}i0 denotes the exchange current density of the electrochemical reaction, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{a}$$\end{document}αa and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{c}$$\end{document}αc refers to the anodic and cathodic transfer coefficient of lithium intercalation and deintercalation reaction, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F$$\end{document}F is faraday’s constant. The macroscopic stress tensor \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\boldsymbol{\sigma }}}$$\end{document}σ and the vector of volumetric forces \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\bf{b}}}$$\end{document}b of the stress field at the electrode scale should satisfy Eq. (8) when inertia effects are neglected.7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\nabla \cdot {{\boldsymbol{\sigma }}}+{{\bf{b}}}={{\bf{0}}}$$\end{document}∇⋅σ+b=0 The constitutive equation at the electrode level is given in Eq. (9)8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{ij}={C}_{ijkl}:({\varepsilon }_{tot}-{{f}}_{vol}({c}_{s}){\delta }_{kl}-{\varepsilon }_{p}-{\varepsilon }_{th})$$\end{document}σij=Cijkl:(εtot−fvol(cs)δkl−εp−εth)where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{{{\rm{ijkl}}}}$$\end{document}Cijkl is the effective modulus of elasticity, represents the coupling of the stress component in the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\rm{i}}},{{\rm{j}}}$$\end{document}i,j direction to the strain component in the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\rm{k}}},{{\rm{l}}}$$\end{document}k,l direction, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon }_{{{\rm{tot}}}}$$\end{document}εtot is the total strain, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{{vol}}({c}_{s})$$\end{document}fvol(cs) denotes the chemical strain due to concentration change, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\delta }_{{{\rm{kl}}}}$$\end{document}δkl is Kronecker delta, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon }_{{{\rm{p}}}}$$\end{document}εp denotes plastic strain and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon }_{{{\rm{th}}}}$$\end{document}εth denotes thermal strain. Electrochemical reactions inherently cause temperature increases due to reversible entropy changes during Li+ insertion at the interface, irreversible joule heating effects caused by ionic/electronic resistance, and activation polarization losses. Intrinsic conductivity and diffusion coefficient characterize thermal effects via the Arrhenius equation. These parameters exhibit non-monotonic concentration due to ion pair competition and solvation effects, and are embedded as experimentally calibrated reference values \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varGamma }_{ref}({c})$$\end{document}Γref(c)42. For pore-scale transport limitations, the Bruggeman coefficient is used for correction. The unified expression for multiphysics coupling is integrated into the following form.9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varGamma }^{eff}({c},{T},\varepsilon ){=}{\varGamma }_{ref}({c}){{\varepsilon }}^{brug}exp\left[{-}\frac{{{E}}_{{a}}}{{R}}(\frac{1}{{{T}}_{ref}}-\frac{1}{T})\right]$$\end{document}Γeff(c,T,ε)=Γref(c)εbrugexp−EaR(1Tref−1T)Where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varGamma \in \{{\sigma },{D}\}$$\end{document}Γ∈{σ,D}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$brug$$\end{document}brug is the Bruggeman exponent, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varepsilon$$\end{document}ε denotes porosity. Specific parameter expressions are detailed in Supplementary Information (Supplementary Equation 3). Sensitivity analysis of electrochemical-mechanical coupling model The P2D model of LIBs contains a large number of electrochemical, mechanical, geometric parameters with practical significance. These parameters exhibit distinct sensitivities to output responses (e.g., terminal voltage and diffusion-induced stress, i.e., DIS) under varying operating conditions, while most parameters exert negligible effects on specific outputs. To identify highly sensitive parameters across conditions and enhance identification efficiency, a local sensitivity analysis is conducted for 23 parameters prior to model parameter optimization. Given the coupled electrochemical-mechanical interactions and the complexity introduced by multidimensional parameters, local sensitivity analysis is selected due to its suitability for high-dimensional parameter spaces and computationally intensive models. This approach enables rapid screening of parameters with high impact. The normalized sensitivity coefficients are used to quantitatively analyze the sensitivity of each parameter to the terminal voltage and DIS. The calculation formulas are shown in Eqs. (8) and (9).10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S{I}_{v}=\sqrt{\frac{1}{N}\sum {(V-{V}_{0})}^{2}}$$\end{document}SIv=1N ∑(V−V0)211\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S{I}_{DIS}=\sqrt{\frac{1}{N}\sum {(S-{S}_{0})}^{2}}$$\end{document}SIDIS=1N ∑(S−S0)2 The parameters of the P2D model can be classified into the following categories based on their physical meanings and reaction mechanisms: structural parameters, transport parameters, Kinetic parameters, and material property parameters. Structural parameters, including anode and cathode electrode thickness, separator thickness, radius of active particles, electrode porosity, which directly affect the Li+ transport path and effective reaction area. Transmission parameters include two components: one is the liquid-phase electrolyte diffusion coefficient, electrolyte conductivity, Li+ transfer number, and the other is the solid-phase particle diffusion coefficient, electronic conductivity, etc., dominating the Li+ mass transfer process in the liquid and solid phases. Kinetic parameters include reaction rate constant, electric double-layer capacitance and charge transfer coefficient, which control the electrochemical reaction rate at the electrode-electrolyte interface. The mechanical parameters of material properties contain three aspects: the first is the density, Young’s modulus and Poisson’s ratio of the active particles, the second is the Young’s modulus, density and Poisson’s ratio of the composite electrode as a whole, the third is the Li+ embedding induced volumetric strain function and the partial molar volume. The parameter superscripts + and - represent the cathode and anode, respectively, while the subscripts s and e represent the solid-phase electrode and electrolyte, respectively. The parameter ranges given in Table 1 are the initial value ranges of the parameters; some of the parameter values will be changed as charging and discharging proceeds, controlled by the governing equations in Supplementary Information (Supplementary Equation 1 and Supplementary Equation 2).Table 1The specific parameters of batteries11,46,47CategoryParameterUnitDescriptionBoundaryStructural parameters\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{L}}}}^{+}$$\end{document}L+μmCathode thickness40–79\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{L}}}}^{-}$$\end{document}L−μmAnode thickness40–79\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{L}}}}_{{{\rm{s}}}}$$\end{document}LsμmSeparator thickness10–30\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\varepsilon }}}}_{{{\rm{s}}}}^{+}$$\end{document}εs+/Cathode active material volume fraction0.35–0.5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\varepsilon }}}}_{{{\rm{s}}}}^{-}$$\end{document}εs−/Anode active material volume fraction0.4–0.5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\varepsilon }}}}_{{{\rm{e}}}}^{+}$$\end{document}εe+/Cathode electrolyte volume fraction0.27–0.45\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\varepsilon }}}}_{{{\rm{e}}}}^{-}$$\end{document}εe−/Anode electrolyte volume fraction0.26–0.5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\varepsilon }}}}_{{{\rm{e}}}}^{{{\rm{s}}}}$$\end{document}εes/Separator electrolyte volume fraction0.4–0.55\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{r}}}}_{{{\rm{p}}}}^{+}$$\end{document}rp+μmCathode particle radius1–11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{r}}}}_{{{\rm{p}}}}^{-}$$\end{document}rp−μmAnode particle radius1–11Transport parameters\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\sigma }}}}_{{{\rm{s}}}}^{+}$$\end{document}σs+S·m-1Cathode electrode conductivity36–185\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\sigma }}}}_{{{\rm{s}}}}^{-}$$\end{document}σs−S·m-1Anode electrode conductivity1–104\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{D}}}}_{{{\rm{s}}}}^{+}$$\end{document}Ds+10-14m2·s-1Cathode solid diffusion1–10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{D}}}}_{{{\rm{s}}}}^{-}$$\end{document}Ds−10-14m2·s-1Anode solid diffusion1–10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{D}}}}_{{{\rm{e}}}}$$\end{document}De10-10m2·s-1Electrolyte diffusion1.5–4.5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{t}}}}_{0}$$\end{document}t0/Li+ transfer number0.25–0.43\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{c}}}}_{{{\rm{e}}},0}$$\end{document}ce,0103mol·m-3Initial electrolyte concentration1–1.2Kinetic parameters\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\kappa }}}}^{+}$$\end{document}κ+10-11m·s-1Cathode reaction rate coefficient1–10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\kappa }}}}^{-}$$\end{document}κ−10-11m·s-1Anode reaction rate coefficient1–10Mechanical parameters\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\rm{E}}}$$\end{document}EGPaAnode particle Young’s modulus10–40\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\rm{\nu }}}$$\end{document}ν/Anode particle Poisson’s ratio0.2–0.35\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\Omega }^{-}$$\end{document}Ω−10-6m3·mol-1Anode partial molar volume4.9–7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\sigma }}}}_{{{\rm{ys}}}}$$\end{document}σysGPaAnode electrode yield stress0.3–1Other parameters\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{c}}}}_{{{\rm{s}}},\max }^{+}$$\end{document}cs,max+104mol·m-3Cathode maximum lithium concentration4.1–5.2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{c}}}}_{{{\rm{s}}},\max }^{-}$$\end{document}cs,max−104mol·m-3Anode maximum lithium concentration2.8–3.3 Table 1The specific parameters of batteries11,46,47CategoryParameterUnitDescriptionBoundaryStructural parameters\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{L}}}}^{+}$$\end{document}L+μmCathode thickness40–79\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{L}}}}^{-}$$\end{document}L−μmAnode thickness40–79\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{L}}}}_{{{\rm{s}}}}$$\end{document}LsμmSeparator thickness10–30\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\varepsilon }}}}_{{{\rm{s}}}}^{+}$$\end{document}εs+/Cathode active material volume fraction0.35–0.5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\varepsilon }}}}_{{{\rm{s}}}}^{-}$$\end{document}εs−/Anode active material volume fraction0.4–0.5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\varepsilon }}}}_{{{\rm{e}}}}^{+}$$\end{document}εe+/Cathode electrolyte volume fraction0.27–0.45\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\varepsilon }}}}_{{{\rm{e}}}}^{-}$$\end{document}εe−/Anode electrolyte volume fraction0.26–0.5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\varepsilon }}}}_{{{\rm{e}}}}^{{{\rm{s}}}}$$\end{document}εes/Separator electrolyte volume fraction0.4–0.55\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{r}}}}_{{{\rm{p}}}}^{+}$$\end{document}rp+μmCathode particle radius1–11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{r}}}}_{{{\rm{p}}}}^{-}$$\end{document}rp−μmAnode particle radius1–11Transport parameters\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\sigma }}}}_{{{\rm{s}}}}^{+}$$\end{document}σs+S·m-1Cathode electrode conductivity36–185\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\sigma }}}}_{{{\rm{s}}}}^{-}$$\end{document}σs−S·m-1Anode electrode conductivity1–104\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{D}}}}_{{{\rm{s}}}}^{+}$$\end{document}Ds+10-14m2·s-1Cathode solid diffusion1–10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{D}}}}_{{{\rm{s}}}}^{-}$$\end{document}Ds−10-14m2·s-1Anode solid diffusion1–10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{D}}}}_{{{\rm{e}}}}$$\end{document}De10-10m2·s-1Electrolyte diffusion1.5–4.5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{t}}}}_{0}$$\end{document}t0/Li+ transfer number0.25–0.43\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{c}}}}_{{{\rm{e}}},0}$$\end{document}ce,0103mol·m-3Initial electrolyte concentration1–1.2Kinetic parameters\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\kappa }}}}^{+}$$\end{document}κ+10-11m·s-1Cathode reaction rate coefficient1–10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\kappa }}}}^{-}$$\end{document}κ−10-11m·s-1Anode reaction rate coefficient1–10Mechanical parameters\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\rm{E}}}$$\end{document}EGPaAnode particle Young’s modulus10–40\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\rm{\nu }}}$$\end{document}ν/Anode particle Poisson’s ratio0.2–0.35\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\Omega }^{-}$$\end{document}Ω−10-6m3·mol-1Anode partial molar volume4.9–7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\sigma }}}}_{{{\rm{ys}}}}$$\end{document}σysGPaAnode electrode yield stress0.3–1Other parameters\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{c}}}}_{{{\rm{s}}},\max }^{+}$$\end{document}cs,max+104mol·m-3Cathode maximum lithium concentration4.1–5.2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{c}}}}_{{{\rm{s}}},\max }^{-}$$\end{document}cs,max−104mol·m-3Anode maximum lithium concentration2.8–3.3 The specific parameters of batteries11,46,47 The terminal voltage, as an external characteristic of the battery, can be measured directly, mainly reflecting thermodynamic equilibrium and kinetic polarization. However, the response to mechanical degradation is lagging or insensitive since stress concentration may occur earlier than voltage decay. Moreover, in the electrochemical-mechanical coupling model, the mechanical response is highly nonlinearly coupled with the electrochemical process. Traditional voltage metrics cannot directly reflect the sensitivity of mechanical parameters. Therefore, a cross-scale and cross-physics metric that can reflect the mechanical characteristics is attempted to be identified. Hydrostatic stress as a direct mechanical characterization of volume change that quantifies the hydrostatic pressure (isotropic stress component) to which the material is subjected. It is highly correlated with both the concentration gradient of the electrochemical field and the material expansion/contraction of the mechanical field. By analyzing its sensitivity to parameters, it is possible to identify key parameters that are prone to stress concentration or plastic deformation, thus providing new perspectives for the subsequent prediction of mechanical failure risk and optimization of electrode material selection or structural design. To ensure the accuracy and reliability of the simulation results, a mesh independence verification is rigorously conducted. This verification confirms that the results for both the global cell voltage and the local current density at the electrode scale are independent of the mesh size, thereby establishing confidence in the predictive capability of the model prior to the formal performance investigation. The comparison results between coarse and fine meshes can be found in the Supplementary Information (Supplementary Fig. 6). Coupled model parameter identification Charging-discharging tests were performed to measure the strain induced in the battery due to expansion. The simulation model developed is desired to accurately predict the voltage while fitting the stress-strain during charging or discharging of the battery. The inherent conflict between voltage and strain as competing objectives creates a trade-off dilemma, where simultaneous optimization of both parameters is elusive. To resolve this multi-objective challenge, a Pareto improvement strategy is adopted, which seeks incremental adjustments to approach optimality without sacrificing either objective. Within this framework, a solution set A is defined to dominate another set B if A is non-inferior in all objectives and strictly superior in at least one, thereby establishing a hierarchy of Pareto-optimal solutions. The set of solutions that are not dominated by any other solution constitutes a Pareto frontier that characterizes the optimal situation after a trade-off between objectives. The mathematical definition and formula of Pareto optimality is given in Eq. (10).12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{array}{cc}\min ({f}_{1}(x){,}{f}_{2}(x),\ldots ,{f}_{m}(x)) & subject\,to\,x\in X\end{array}$$\end{document}min(f1(x),f2(x),…,fm(x))subjecttox∈Xwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m$$\end{document}m is the number of objective functions of the multi-objective optimisation problem, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$X$$\end{document}X is the set of feasible solutions and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x$$\end{document}x is the decision variable. A solution \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${x}^{* }\in X$$\end{document}x*∈X is Pareto optimal, if and only if there exists no other solution \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x\in X$$\end{document}x∈X such that Eq. (11) holds and there exists at least one \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$j\in \{{\mathrm{1,2}}\ldots ,m\}$$\end{document}j∈{1, 2…,m} that satisfies Eq. (12).13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\forall i\in \{1,2,\ldots ,m\},\,{f}_{i}(x)\le {f}_{i}({x}{\ast })$$\end{document}∀i∈{1,2,…,m},fi(x)≤fi(x*)14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{j}(x) < {f}_{j}({x}{\ast })$$\end{document}fj(x)