Interfacial excess charge dynamics as a quantitative descriptor to understand lithium dendrite growth

Abstract

Unraveling the origins of Li dendrite growth remains a critical challenge for lithium metal batteries. Existing studies often treat diffusion-reaction mismatches or electrode surface excess charge in isolation, overlooking the coupled interfacial processes at the nanoscale solid-liquid interface that govern Li dendrite growth. Herein, we introduce the interfacial excess charge distribution factor, a quantitative descriptor that captures the dynamic equilibrium by integrating the contributions of electrode surface excess charge, charge depletion rate and solvation chemistry. We further propose an electrochemical method to monitor Li dendrite growth rates, confirming that electrolytes with rapid excess charge redistribution ability favor planar Li plating. Our findings reveal that the depletion rate of interfacial excess charge governed by Li⁺-dipole-anion interactions, and the degree of electrode excess charge modulated by solvation structures and cation screening, competitively determine the interfacial excess charge dynamics. Guided by these insights, we designed an electrolyte with optimized excess charge redistribution capability, which suppresses dendrite formation and enables stable cycling in ~6.4 Ah lithium metal batteries. This work provides mechanistic insight into how competitive interfacial processes dictate Li dendrite growth.

Tracing the origins of lithium dendrite growth remains a critical challenge for lithium metal batteries. Here, the authors introduce an interfacial excess charge distribution factor to understand lithium dendrite growth, guiding the design of an electrolyte that suppresses Li dendrites in practical pouch cells.

Introduction

As the demand for higher energy density grows in battery technology, lithium-based energy storage mechanisms need to rely on the reversible electrochemical Li plating/stripping¹. The electroplating mechanism promotes the development of lithium metal batteries (LMBs), whose specific gravimetric energy can exceed 600 Wh kg^−1 2–4. However, Li electroplating involves complex steps, including bulk Li⁺ diffusion, desolvation, charge transfer, and eventually electro-crystallization of neutral Li atoms into metallic Li⁵. The aggregation of Li atoms typically gives rise to macroscopically structured deposits with fractal-like morphologies⁶. The morphological instability becomes more complex when dealing with chemically reactive metallic lithium in battery environment^7,8. Li dendrites not only penetrate the solid electrolyte interphase (SEI) to deplete Li inventory, but their tips also tend to detach during stripping, causing loss of conductivity to form problematic dead Li (i.e., metallic Li, which is encapsulated by SEI and electronically disconnected from the current collector)^9–11.

Tracing the origins of dendrite formation during electrochemical Li plating/stripping is therefore important for developing LMBs. Since electroplating occurs exclusively at the electrode/electrolyte interface, the thermodynamic and kinetic characteristics can govern both Li plating behavior and the resulting deposit morphology. From the kinetic perspective, dendrite initiation primarily arises from the rate mismatch between Li⁺ diffusion in the bulk electrolyte and Li⁺ consumption at the solid-liquid interface^12,13. This mismatch is quantitatively captured by models such as Sand’s time^13,14, which predicts the onset of interfacial Li⁺ depletion under a given current density. Furthermore, Bazant confirmed that applied current densities exceeding the critical threshold induce localized Li⁺ depletion during electroplating^13,15–17. Thus, kinetic insights have enabled new strategies to suppress dendrite growth, such as increasing electrolyte concentration¹⁸, optimizing charging/discharging protocols¹⁹, enhancing Li⁺ transport through the SEI²⁰, and reducing interfacial nucleation overpotentials⁵. However, Li electroplating is a complex process occurring at the solid-liquid interface, governed by electrode surface properties, solvent-anion clusters in the electric double layer (EDL), and electroplating kinetics^21–23. Yet, thermodynamic approaches to reveal these effects remain limited. This is because understanding dendrite formation from the thermodynamic perspective is challenged by the lack of suitable descriptors capable of capturing the complexity beyond simple directional tendencies.

Recently, electrode surface excess charge at the electrode/electrolyte interface, arising from mismatches in electronic energy levels between electrodes and electrolytes, has emerged as a promising descriptor to elucidate the thermodynamic origins of Li dendrites²⁴. Negative excess charge accumulates whenever the electroplating potential (E_plating) is more negative than the electrode’s potential of zero charge (E_PZC). As illustrated in Fig. 1a, metals such as Cu, Ag, and Ni typically operate in the regime of positive surface charge, repelling cations and promoting uniform deposition. In contrast, metals like Li, Na, K, and Mg operate at potentials significantly negative relative to their respective E_PZC. The generation of different interfacial excess charges among metals at the electrode/electrolyte interface depends on the electrode characteristics (e.g., crystallographic orientation, work function, alloy composition) and electrolyte properties (e.g., solvation structure, solvent reorganization, and EDL configuration). While the electrochemical black-box nature of batteries often obscures direct observation, quantitative descriptors that link internal thermodynamic and kinetic states to macroscopic morphology are needed to reframe and understand these hidden dynamics.

Fig. 1. Excess charge of the Li/electrolyte interface across distinct electrolytes.

a Demonstration of the relationship between surface tension and the potential differences between E_PZC and E_plating for various electroplating pairs (Mⁿ⁺/M), which determine the electroplating behavior. b Correlation between electrode excess charge (the difference between E_PZC and E_plating) and CE_Li for different electrolytes. Data are presented as mean values ± S.D., the labels (3/3) specify the numbers of independent experiments for the x- and y-axis data, respectively. c Schematic of solvation structures in the inner Helmholtz plane (IHP), outer Helmholtz plane (OHP), and diffusion layer, and the related electrode surface charge state. μ_e,m and μ_e,s correspond to the Fermi levels of the metal electrode and the electrolyte, respectively, while μ⁰_e,m, and μ⁰_e,s represent their respective standard states. Symbols: circles (Li⁺), ellipses (solvents), and diamonds (anions). d Percentage of different Li⁺-solvent-anion clusters for electrolyte D/F/J according to CMD simulations. e Prediction of atomic charge at Li/electrolyte interface by using a machine learning model; the configurations of the solvents (DEE, DME) and Li salt (LiFSI) are present on the right. f Correlation between the calculated charge from the machine learning model and practical CE_Li, which aligns with the results in (b). Excess charge distribution along the x-y axes for Li atom layers at the Li/electrolyte interface in the machine learning model for g electrolyte D, h electrolyte F, and i electrolyte J.

Current studies rely on theoretical calculations and molecular dynamics (MD) simulations to reveal the role of interfacial excess charge^20,25–28. For instance, Filhol et al. proposed that the accumulation of excess charge leads to a reduction in the electrode surface tension according to the Lippmann equation^29,30. From a qualitative thermodynamic perspective, this reduction drives the electrode surface to spontaneously expand its area to mitigate surface tension, and the resulting increase in surface area is typically interpreted as Li dendrite growth³¹. However, while theoretical calculations and MD simulations can predict this tendency for surface area enlargement, they are unable to quantify the relationship between the morphological change and interfacial excess charge. Consequently, the thermodynamic influence of interfacial excess charge on Li dendrite growth remains insufficiently understood from a quantitative perspective. Robust experimental methodologies capable of quantitatively correlating interfacial excess charge with dendrite formation are therefore needed. However, such experimental investigations are hindered by two significant challenges: (1) experimental methods to track the generation, consumption, and distribution of interfacial excess charge remain underdeveloped; (2) dendrite growth occurring within the sealed cell makes it challenging to achieve meaningful insights through post-mortem morphological analyses alone. Hence, developing experimental methodologies that quantitatively evaluate interfacial excess charge dynamics and correlate these findings with dendrite growth will advance our understanding.

To address these challenges, we introduce a descriptor, the interfacial excess charge distribution factor, defined as the exchange current density ( J₀) normalized by the net interfacial excess charge. Physically expressed in units of frequency (Hz), this parameter integrates interfacial excess charge, electrolyte solvation structures, and electroplating reaction kinetics into a single metric that captures the dynamic equilibrium at the nanoscale solid-liquid interface. We experimentally quantify this factor by measuring the E_plating, E_PZC, and J₀ across diverse electrolytes, while validating and predicting the atomistic charge distributions in large-scale Li/electrolyte systems via the developed machine learning (ML) model. Additionally, to correlate this descriptor with the morphological evolution of Li plating, we developed an electrochemical method to evaluate the growth rates of electrochemically active surface area (ECSA) for Li plating. We confirm that higher ECSA growth rates directly correspond to more severe dendrite formation, which aligns with the fractal morphology analyzed by small-angle neutron scattering (SANS), synchrotron X-ray computed tomography (XCT), and phase-field simulations. Our analysis reveals that the redistribution dynamics of interfacial excess charge are governed by the competitive interplay between cation screening effects and the strength of Li⁺-dipole interactions. Guided by this mechanistic insight, we demonstrate the practical application of an electrolyte with rapid interfacial excess charge redistribution capability. By using this electrolyte, a 6.4 Ah LMB achieves 515 Wh kg⁻¹ under a low negative-capacity to positive-capacity (N/P) ratio (0.80) and lean electrolyte injection ratio of 1.00 g Ah⁻¹, delivering improved cycling stability compared to baseline formulations. Ultimately, deciphering these competitive interfacial dynamics offers a critical mechanistic pathway to understand and control Li dendrite growth.

Results

To investigate the influence of solvation characteristics on electrode surface excess charge, we examined 12 electrolytes (denoted as A to L, detailed formulations and abbreviations provided in Supplementary Tables 1 and 2) with varying solvation structures. These electrolytes span diverse solvation regimes, including diluted electrolytes (DE), diluted weak solvation electrolytes (WSE), high-concentration electrolytes (HCE), high-concentration weak solvation electrolytes (HWSE), and localized high-concentration electrolytes (LHCE). Their physicochemical properties, including conductivity, viscosity, and electrochemical stability windows, are detailed in Supplementary Fig. 1. To decouple the complex influence of electrolyte additives on the interphase components, we employed a simplified formulation strategy using structurally analogous solvents and a single salt (LiFSI). By modulating the solvation structure solely through concentration changes and structurally analogous solvent substitution, and by applying the designed preconditioning procedures (in the “Methods” section) to remove the native oxides on Li foil, we minimized the variations in the initial SEI, as supported by X-ray photoelectron spectroscopy (XPS) and time-of-flight secondary ion mass spectrometry (TOF-SIMS) analyses (Supplementary Figs. 2 and 3). These electrolytes were used to correlate the Coulombic efficiency of Li plating/stripping (CE_Li) with electrode surface excess charge, as illustrated in Fig. 1b. Detailed electrochemical protocols for determining CE_Li, E_plating, and E_PZC are provided in the “Methods” section, with data and statistical confidence intervals for each electrolyte provided in Supplementary Figs. 4–18 and Table 3. For electrolytes A to D, categorized as DE, we observe that as the electrode excess charge decreases, the CE_Li increases; this trend is also observed in the WSE regime. However, for high-concentration electrolytes, including the HCE/LHCE group (G/J/K/L) and the HWSE group (H/I), the electrode excess charge increases with Li⁺ concentration, which is accompanied by an increase in CE_Li. Figure 1c outlines the mechanism by which EDL solvation structures influence electrode excess charge. When the Fermi energy levels of a partial-desolvated solvation cluster align with those of a metal electrode, the potential of the solvation cluster being higher than that of the metal electrode will lead to electron transfer from the solvation structure (high potential) to the electrode (low potential)^32,33. As a result, excess charge accumulates at the electrode surface, which highlights the role of solvation structure in influencing the state of electrode excess charge.

Classic MD (CMD) simulations were performed for electrolytes D (DE), F (WSE), and J (HCE) at the Li/electrolyte interface (Fig. 1d, Supplementary Figs. 19 and 20 and Supplementary Data 1). The results reveal that as the Li⁺-dipole interaction weakens, whether through  the use of low-permittivity solvent or high salt concentrations, solvents tend to be repelled to form Li⁺-dipole-anion clusters, and the proportion of Li⁺-dipole pairs decreases. In WSE, the weak Li⁺-dipole interaction facilitates the entry of more negatively charged anions into EDL, lowering the potential of partial-desolvated solvation clusters, in turn, reducing interfacial excess charge. However, in HCE, despite the increased anion population in EDL, the rise in the cation concentration outweighs the potential changes. Consequently, with higher concentrations of Li⁺, the HCE, HWSE, and LHCE all show an increasing trend in electrode excess charge. To further assess atomic charge distribution in large-scale systems, which classical force fields cannot capture, we developed an ML model (details in “Methods” and Supplementary Note 1/Table 4) capable of predicting the atomic charge of the Li electrode³⁴. Fig. 1e presents the snapshots of the atomic charge distribution across the entire construction. Owing to the low-dissociation ability of its solvent, electrolyte F displays lower and more dispersed charge density than that of electrolyte D. The atomic charge densities of Li electrodes in three electrolytes are calculated to reveal the direct correlation with measured CE_Li in Fig. 1f, which is consistent with experimental results. Furthermore, Fig. 1g–i shows the charge density distribution on the Li electrode after simulating for 5 ns. In electrolyte D, the charge distribution remains relatively uniform. In contrast, electrolyte F shows a slight reduction in charge coverage, due to the decrease in interfacial excess charge. In electrolyte J (HCE), the uniformity of surface charge distribution diminishes, giving rise to distinct regions of localized charge accumulation. This heterogeneity is attributed to the high local cation concentration, which leads to an overall increase in interfacial charge load.

Figure 2a illustrates the ML-derived charge density distribution model at the electrode/electrolyte interface, presenting the excess charge accumulation on the electrode against the corresponding solvation structures in DE, WSE, and HCE. In DE, Li⁺-solvent interactions dominate the interface to promote excess charge accumulation on the electrode. In contrast, WSE weakens these interactions, which enables anions to locally participate in Li⁺-solvent-anion coordination. This interaction lowers the potential of solvation clusters, effectively reducing the excess countercharge. As Li⁺ concentration increases, the rising cation density near the interface leads to an increase in charge density, necessitating the corresponding buildup of countercharge on the electrode to maintain electroneutrality. In HCE, Li⁺-anion coordination predominates with less solvent involvement; localized variations of anion and solvent in solvation coordination contribute to the uneven distribution of excess charge, but with a higher charge density level. However, evaluating Li dendrite growth solely based on the magnitude of interfacial excess charge fails to yield a consistent trend. The divergent behaviors observed between DE/WSE and HCE/HWSE/LHCE suggest that while electrode excess charge serves as a meaningful descriptor in diluted electrolytes, it does not fully capture the charge dynamics in HCE.

Fig. 2. Quantitative descriptor for the interfacial excess charge redistribution dynamics.

a Schematic of Li⁺-dipole-anion interactions in the electric double layer (EDL) and corresponding electrode excess charge states for different electrolytes. Symbols: circles (Li⁺), ellipses (solvents), and diamonds (anions). b Schematic illustrating how interfacial excess charge redistribution dynamics govern the interfacial properties for Li plating. c Concept of the interfacial excess charge distribution factor, determined by electrode excess charge, exchange current density, and cation screening effects. d Li⁺ depletion zones along the z-axis (perpendicular to electrode/electrolyte interface) for electrolyte D/F/J, derived from the results of the machine learning model. Top-view spatial Li⁺ distribution in the x-y plane for e electrolyte D, f electrolyte F, and g electrolyte J. The green spheres in (e–g) represent the Li-ions at the electrolyte/electrode interface. In situ electrochemical ATR-FTIR peaks for electrolyte D (h), electrolyte F (i), and electrolyte K ( j). The top panel of each figure illustrates the peaks assigned to free solvents and coordinated Li⁺-solvent. In electrolyte K, the peak intensity for Li⁺-solvent is lower than in electrolyte D due to the addition of the diluent. The chronopotentiometry with a current density of 2.0 mA cm⁻² for 1800 s was employed for monitoring the desolvation process on the Cu mesh surface with a home-built electrochemical cell. The curves in (h–j) represent the derivatives. k Comparison of exchange current densities across electrolytes. Data are presented as mean values ± S.D., and the labels specify the numbers of independent experiments. Correlation between the calculated interfacial excess charge distribution factor and the corresponding CE_Li for electrolytes: l DE/WSE, m HWSE, and n HCE/LHCE regimes. DE, WSE, HWSE, HCE, and LHCE mean diluted electrolyte, weak solvation electrolyte, high-concentration weak solvation electrolyte, high-concentration electrolyte, and localized high-concentration electrolyte, respectively.

Previous studies have attributed the improvement in CE_Li within WSE to anion involvement in solvation structures, which is thought to promote the formation of inorganic-rich SEI³⁵. However, the comparison between electrolyte F and G highlights the notable discrepancy: despite electrolyte G being an HCE with higher anion participation in solvation, which enables a stronger potential for inorganic SEI formation, its average CE_Li of 98.8% remains lower than that of electrolyte F (99.2%). This observation suggests that relying on the extent of anion involvement in solvation alone does not fully explain the variation in CE_Li across different electrolytes. The underlying cause can be linked to interfacial charge distribution dynamics. In DE, electrode excess charge plays a pivotal role in determining interfacial charge distribution uniformity, and the weaker Li⁺-solvent interaction reduces excess charge accumulation. However, in HCE, the EDL is characterized by a higher cation concentration, where the number of anion coordination sites far exceeds that of solvents in the solvation structure. Under such conditions, the driving forces governing EDL charge balance and uniformity are primarily dictated by Li⁺-anion interactions rather than Li⁺-solvent interactions. Therefore, static electrode excess charge alone is an insufficient metric to explain how solvation structures shape interfacial charge dynamics.

To bridge the gap in understanding interfacial excess charge behavior, we propose a transition from a static to a dynamic perspective, one that accounts for both electrochemical consumption of interfacial charge and its spatiotemporal redistribution. The desolvation rate of Li⁺, governed by the specific solvation environment, directly dictates the consumption rate. Meanwhile, the combined effect of electrode excess charge and cation accumulation driven by high Li⁺ concentration introduces further complexity to spatiotemporal redistribution. Figure 2b, c illustrate this dynamic perspective through the concept of interfacial charge redistribution dynamics, which comprise three interdependent components: (1) electrode surface excess charge; (2) the charge depletion rate via interfacial electrochemical reactions; (3) cation screening that offsets local charge imbalances. Among these, the depletion rate reflects electrochemical kinetics and can be quantitatively described by J₀. In contrast, both electrode excess charge and cationic screening emerge from the electrostatic response of the interface. Within this framework, we introduce the interfacial excess charge distribution factor ($f$), defined as J₀ divided by the net surface charge (${\mathrm{\sigma }}_{\mathrm{net}}$) in Expression 1. This descriptor reflects the interface’s dynamic capability to neutralize excess charge imbalances via electrochemical reaction. Under Li plating conditions, the electrode surface accumulates negative electrode excess charge (${\mathrm{\sigma }}_{\mathrm{surface}}$) driven by the thermodynamic potential bias, whereas the EDL accumulates a dense population of positively charged cations (${\mathrm{\sigma }}_{\mathrm{screen}}$) that attempts to neutralize this charge. Since these two populations possess opposite polarities, the dense cation layer acts to screen the electrode’s charge stress. Thus, the difference between ${\mathrm{\sigma }}_{\mathrm{surface}}$ and ${\mathrm{\sigma }}_{\mathrm{screen}}$ quantifies the uncompensated charge imbalance remaining at the interface:

$$f=\frac{{\mathrm{J}}_0}{{\mathrm{\sigma }}_{\mathrm{net}}}=\frac{{\mathrm{J}}_0}{{\mathrm{\sigma }}_{\mathrm{surface}}-{\mathrm{\sigma }}_{\mathrm{screen}}}$$ 1

where J₀ denotes the exchange current density, ${\mathrm{\sigma }}_{\mathrm{excess}}$ represents the electrode excess charge density, and ${\mathrm{\sigma }}_{\mathrm{screen}}$ corresponds to the countercharge contributed by interfacial cations. ${\mathrm{\sigma }}_{\mathrm{excess}}$ can be calculated through EDL capacitance (${\mathrm{C}}_{\mathrm{dl}}$), while the cation screening effect is expressed as a function of electrolyte concentration (${\mathrm{c}}_0$), as detailed in the “Methods” section and Supplementary Note 2. Based on the described relationships, the overall relationship of the interfacial charge distribution factor can be generalized in Expression 2:

$$f=\frac{{\mathrm{J}}_0}{\left({\mathrm{E}}_{\mathrm{PZC}}-{\mathrm{E}}_{\mathrm{plating}}\right)*{\mathrm{C}}_{\mathrm{dl}}-\mathrm{\Upsilon }{{*\mathrm{c}}_0}^{0.5}}$$ 2

Here, $\mathrm{\Upsilon }$ and ${{\mathrm{c}}_0}^{0.5}$ represent the magnitude and scaling behavior of the concentration-dependent screening effects, respectively. Since J₀ represents charge flow per second per unit area and ${\mathrm{\sigma }}_{\mathrm{net}}$ represents charge per unit area, the dimensional analysis confirms that $f$ carries the unit of frequency (Supplementary Table 5), demonstrating the characteristic rate of interfacial charge equilibration. Accordingly, a higher $f$ suggests a faster redistribution frequency and thus more homogeneous interfacial charge balances.

Figure 2d illustrates how Li⁺ depletion zones³⁶ evolve across distinct electrolytes, offering insights into the influence of solvation structures on interfacial reaction kinetics. CMD simulations reveal that transitioning from electrolyte D (DE) to electrolyte F (WSE) results in a notable increase in the Li⁺ depletion zone along the z-axis direction. This effect arises primarily from the weakened Li⁺-dipole interactions in WSE, which effectively reduces the desolvation barrier for Li⁺ (Supplementary Table 6/Note 3), thereby accelerating interfacial reaction rates. Conversely, increasing the salt concentration to formulate electrolyte J (HCE) compresses the depletion zone. This compression is driven by stronger Li⁺-anion coordination in HCE, which increases the energy penalty for Li⁺ desolvation (Supplementary Table 6/Note 3), thus retarding the interfacial consumption of excess charges. Additionally, the dense accumulation of Li⁺ within the EDL intensifies this compression due to the increased cation concentration. Further insights emerge from Fig. 2e–g, which present spatial distributions of interfacial Li⁺, elucidating the role of cation screening in compensating interfacial charge imbalance. The frequency annotations at the boundaries of each figure illustrate Li⁺ distributions along the x-y plane. In the DE and WSE regime, characterized by lower bulk concentrations, Li⁺ exhibits relatively nonuniform distributions across the interface. In contrast, HCE electrolytes exhibit densely packed and homogeneous Li⁺ arrangements in EDL, highlighting more effective charge screening and uniform interfacial cation distribution. To provide direct experimental evidence of these interactions, we performed in situ electrochemical attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR). In Fig. 2h–j, the Li⁺-DEE coordination peak emerges earliest in electrolyte F (128 s), followed by electrolyte K (198 s) and electrolyte D (309 s) with Li⁺-DME coordination, confirming that electrolyte F possesses the fastest desolvation kinetics. Notably, electrolyte K reaches its maximum peak intensity at 611 s, attributed to the high cation density that facilitates immediate EDL saturation for the strong screening effect. In contrast, electrolytes D and F require ~1800 s to reach saturation, indicating weaker screening capability. Based on the analysis above, the shift from solvent-separated ion pairs to contact ion pairs alters the E_PZC, while the crowding of cations and anions in HCE regime not only enhances Li⁺-anion interactions but also creates the dense Helmholtz plane, amplifying the screening term. These results demonstrate that solvation chemistry causally dictates both the numerator and the denominator of f.

The depth of depletion zone (z-axis) and the lateral Li⁺ distribution uniformity (x-y plane) support the viability of analyzing the proposed interfacial excess charge distribution factor. Among the three parameters, the degree of electrode excess charge has already been measured (Fig. 1b), and cation screening scales with bulk electrolyte concentration. Thus, comparing f across different electrolytes now hinges on experimentally measuring J₀ (details in the “Methods” section), which reflects the kinetic rate of interfacial charge depletion. Representative current-voltage curves, electrochemical impedance spectroscopy (EIS), calculated J₀, and related confidence intervals for each electrolyte are shown in Supplementary Figs. 4–15 and 21. The corresponding J₀ values calculated for electrolyte A-L are summarized in Fig. 2k. From DE (electrolyte A/B/C/D) to WSE (electrolytes E/F), J₀ demonstrates a clear upward trend, from 2.49 mA cm⁻² of electrolyte A to 6.55 mA cm⁻² of electrolyte E, and even 9.33 mA cm⁻² of electrolyte F, directly correlating with the progressive weakening of Li⁺-dipole interactions³⁵. This increase validates that WSEs facilitate more rapid consumption of interfacial excess charge. However, upon further increasing Li⁺ concentration and reducing the solvation ability of the solvent, J₀ gradually decreases from 8.84 mA cm⁻² of electrolyte G to 6.58 mA cm⁻² of electrolyte I, and drops further to 2.57 mA cm⁻² in electrolyte J, suggesting that the strengthened Li⁺-anion coordination can’t effectively enable fast kinetics. When diluting HCEs with 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropylether (TTE) to generate LHCE (electrolyte K and L), an increase in J₀ is observed, reflecting enhanced interfacial kinetics attributed to the modulation of anion-solvent coordination by the diluent. A single parameter, such as J₀ or surface excess charge, alone fails to explain dendrite suppression across these electrolytes, as high reversibility can stem from either fast kinetics or strong screening. The descriptor f can resolve this paradox by integrating these factors into the proposed metric, simply quantifying the interface’s overall capacity to dissipate charge accumulation that triggers dendrite growth.

Due to inherent experimental challenges in explicitly resolving the exact cation screening effects, we classified electrolytes into distinct categories and strategically omitted minor contributing factors within each class. This approach facilitates a clearer, quantitative comparison of f across various electrolytes. For diluted electrolytes (including DEs and WSEs), the previously discussed cation screening effect is negligible (Supplementary Note 2), leaving the Li⁺-dipole interactions as the predominant influence. Consequently, the general Expression 2 simplifies to Supplementary Expression S3. As depicted in Fig. 2l, taking electrolyte D as the reference baseline, the clear increase in both CE_Li and f is evident across electrolytes A to F. Consistent with prior analyses, weakening Li⁺-dipole interactions in these electrolytes reduces electrode excess charge and elevates J₀. This combined effect enhances the depletion rate of excess charge, thereby diminishing its detrimental impact on the reversibility of Li plating/stripping. In the HWSE group (Fig. 2m), where both cation screening and Li⁺-dipole interactions can determine charge behavior, electrolyte G is designated as the baseline, enabling comparative evaluation of electrolytes H and I (Supplementary Expression S5). While concurrent increases in both electrode excess charge and J₀ are observed across electrolyte G/H/I, a distinct upward trend in f emerges, driven by the weaker Li⁺-dipole interactions. This trend aligns positively with the measured CE_Li, reinforcing the validity of f as the metric that accounts for interfacial dynamics. Finally, in the HCE and LHCE groups (Fig. 2n), CMD simulations reveal minimized Li⁺-dipole interactions, leading to negligible differences in electrode excess charge among electrolytes J/K/L. The interfacial charge balance is dominated by cation screening effects in the EDL. Therefore, Expression 1 can be further simplified to Supplementary Expression S4. By using electrolyte J as the baseline, a consistent positive correlation between f and CE_Li is observed for electrolytes J/K/L. The higher values of f correspond directly to the increased frequencies of interfacial charge redistribution, resulting in more homogeneous dynamic charge distribution and enhanced electrochemical reversibility. These quantitative comparisons establish the proportional relationship between f and CE_Li across DE/WSE, HWSE, and HCE/LHCE regimes.

To establish the quantitative relationship between the interfacial charge distribution factor and Li dendrite growth, it is critical to develop reliable metrics for demonstrating dendrite-growth dynamics. Currently, the analysis of Li dendrite growth relies on localized post-mortem morphology, lacking global and quantifiable parameters for precise comparative analysis. For instance, Hwang et al. proposed utilizing mosaic image analysis via optical microscopy to characterize Li dendrite growth by monitoring the projected surface area of dendrites³⁷. Their findings indicated that the ECSA associated with planar Li plating evolves more slowly compared to dendritic growth. The concept resonates with established methods in electrochemical metallurgy, wherein potentiostatic control is employed to elucidate fundamental electroplating phenomena³⁸. In contrast to galvanostatic electroplating, which provides information on potential fluctuations and overall cell resistance, potentiostatic electroplating affords deeper insights by capturing different stages, including EDL charging, nucleation characteristics, and diffusion-controlled electroplating^39–41. As illustrated in Fig. 3a, the potentiostatic electroplating curve exhibits an initial current peak, followed by the characteristic decline indicative of the transition to the diffusion-limited plating regime. During this stage, reaction kinetics become predominantly diffusion controlled, conforming to the classical Cottrell equation⁴², as easily expressed as Expression 3.

$$\mathrm{i}\propto \frac{\mathrm{ECSA}}{{\mathrm{t}}^{0.5}}$$ 3

Fig. 3. Electrochemical surface area (ECSA) variation during Li plating and its correlation with the interfacial excess charge distribution factor.

a Demonstration of the typical potentiostatic electroplating curve with different stages, showing the current variation for dendrite-free plating and dendrite-growth plating. b Potentiostatic plating curves recorded under different potential biases and time windows, and c the calculated ECSA growth rates. d Correlation between ECSA growth rate and the average CE_Li. Data are presented as mean values ± S.D., and the labels specify the numbers of independent experiments. Quantitative correlation between ECSA growth rate and the interfacial excess charge distribution factor for different electrolyte groups: e DE/WSE, f HWSE, and g HCE/LHCE. h SANS curves of plated Li with an areal capacity of 6.0 mAh cm⁻² at a current density of 0.5 mA cm⁻², and i derived slope values to indicate the different fractal structure characteristics. j Phase-field simulations of Li dendrite growth with random and uniform nucleation sites. Color bars in ( j) show a dimensionless phase-field parameter: 0 corresponds to the liquid electrolyte, 1 corresponds to the solid lithium dendrite, and intermediate values correspond to the transition state. k Calculated standard deviation of each dendrite perimeter in the phase-field simulation in ( j). XCT images for Li plating with areal capacity of 4.0 mAh cm⁻² at current density of 0.5 mA cm⁻² on Cu foil for different electrolytes: l electrolyte A, m electrolyte F, and n electrolyte K.

For conventional inert metals such as Ni or Cu, the current invariably decays over time in the diffusion-limited regime, strictly adhering to ${\mathrm{t}}^{-0.5}$ dependency (represented by the dashed black line in Fig. 3a). However, an atypical curve is observed during potentiostatic electroplating of metallic Li. Specifically, the plating current increases even during the expected diffusion-controlled regime. Results of electrochemical quartz crystal microbalance (Supplementary Fig. 22) indicate the direct correlation between Li deposition mass and charge passed during the plating process, confirming that parasitic side reactions are minimal during this process. Thus, upon analyzing the deviation from the Cottrell relationship, it becomes evident that this rising current can be exclusively attributed to the continuous expansions of ECSA. The baseline for calculating the rising current can neglect the theoretical current density decay (the dashed black line in Fig. 3a) following the inflection point (Supplementary Fig. 23a). Consequently, the magnitude of current increase quantifies Li dendritic growth dynamics, providing a possible metric for dendrite growth. A representative comparison between two distinct current–time responses under potentiostatic electroplating is presented in Fig. 3a (curves 1 and 2). Key points in the diffusion-limited regime are assigned for each curve: (${\mathrm{t}}_1$, ${\mathrm{i}}_{\mathrm{t}1}$) and (${\mathrm{t}}_2$, ${\mathrm{i}}_{\mathrm{t}2}$) denote the onset of ECSA variation, while (${\mathrm{t}}_3$, ${\mathrm{i}}_{\mathrm{t}3}$) and (${\mathrm{t}}_4$, ${\mathrm{i}}_{\mathrm{t}4}$) mark the corresponding endpoints. Within this regime, the variations of the current–time curves reflect the rates of change in ECSA, as assumed ${\mathrm{k}}_1$ and ${\mathrm{k}}_2$ for curves 1 and 2, respectively. As detailed in Supplementary Note 4, the relationship between ${\mathrm{k}}_1$ and ${\mathrm{k}}_2$ can be calculated via following Expression 4.

$$\frac{{\mathrm{k}}_2}{{\mathrm{k}}_1}=\frac{\left(\mathrm{\Delta }{\mathrm{i}}_{\mathrm{t}4-\mathrm{t}2}/\mathrm{\Delta }{\mathrm{t}}_{4-2}\right)*{\mathrm{i}}_{\mathrm{t}1}}{\left(\mathrm{\Delta }{\mathrm{i}}_{\mathrm{t}3-\mathrm{t}1}/\mathrm{\Delta }{\mathrm{t}}_{3-1}\right)*{\mathrm{i}}_{\mathrm{t}2}}*\left(1+\frac{\sqrt{{\mathrm{t}}_1}-\sqrt{{\mathrm{t}}_3}}{\mathrm{\Delta }{\mathrm{t}}_{3-1}*\sqrt{{\mathrm{t}}_3}}\right)-\frac{\sqrt{{\mathrm{t}}_2}-\sqrt{{\mathrm{t}}_4}}{\mathrm{\Delta }{\mathrm{t}}_{4-2}*\sqrt{{\mathrm{t}}_4}}$$ 4

The derived growth rate of ECSA can serve as a reliable descriptor, enabling comparative assessment of dendrite growth rate (k_x) through slope variations observed during the diffusion-limited regime. Figure 3b displays the results of potentiostatic electroplating tests performed on the baseline electrolyte under varied potentials and plating durations, with the corresponding ECSA growth rates summarized in Fig. 3c and Supplementary Table 7/Note 5. Increasing the applied potential enhances electrochemical reaction kinetics, resulting in progressively steeper current slopes in the diffusion-limited regime. Relative to the baseline condition of 75 mV for 0.5 h, the calculated ECSA growth rate increases to 4.9 at 100 mV and further rises to 16.2 at 125 mV, representing multi-fold enhancements relative to the baseline. Moreover, extending plating duration from 0.5 h to 1.0 h at a fixed potential of 100 mV yields an ECSA growth rate of 5.9, closely matching that of 4.9 at 0.5 h. The results of time-window sensitivity analysis indicate that the choice of plating durations has minimal impact on the calculated rate (Supplementary Fig. 23b). To further demonstrate the sensitivity and effectiveness of this methodology, the baseline electrolyte was modified by introducing an additional 5 wt.% fluoroethylene carbonate (FEC). This additive suppresses the slope of the current in the diffusion-limited zone, reducing the corresponding ECSA growth rate to 3.4. The clear contrast provided by this comparative analysis validates that calculating the slope variations in the diffusion-limited regime captures ECSA variations driven by Li dendrite formation, providing a quantitative metric for Li plating dynamics.

Potentiostatic electroplating experiments were conducted for electrolytes A to L by employing electrolyte D as the reference baseline (detailed experimental data presented in Supplementary Figs. 4–15, 24 and Supplementary Table 8). Relative k_x values were calculated and compared across electrolytes. Figure 3d demonstrates the quantitative correlation between CE_Li and k_x. As CE_Li increases, k_x decreases consistently, suggesting that smaller variations in ECSA during Li plating favor a more planar plating morphology. This planar growth reduces the propensity for dendritic formation and mitigates the risk of mechanical detachment at dendrite roots during stripping, a key factor contributing to reduced reversibility. Figure 3e–g further establish the relationships between f and k_x, categorized by electrolyte groups as previously outlined (detailed data shown in Supplementary Table 9). In DE and WSE, the weakening of Li⁺-dipole interaction corresponds to an increase in f, indicative of accelerated interfacial charge redistribution. The k_x decreases from 1.930 (relative to the baseline) of electrolyte A down to 0.297 of electrolyte F. This inverse trend underscores that rapid interfacial charge redistribution substantially reduces the accumulation of excess charge, alleviating the surface energy-driven stress responsible for dendritic Li growth. Similarly, in the HWSE regime, substituting the solvent of electrolyte G with weaker solvating agents such as 1,2-diethoxyethane (DEE) and cyclopentyl methyl ether (CPME) notably elevates f. This increase directly coincides with a marked reduction in ECSA growth rates, from approximately 0.462 to 0.297. The consistent inverse relationship between f and k_x is also evident in the HCE and LHCE groups. These comparisons confirm that the rate of interfacial excess charge redistribution influences the uniformity of charge distribution at the electrode/electrolyte interface. Electrolytes exhibiting faster redistribution dynamics reduce the persistence of interfacial excess charge, thereby minimizing the surface energy burden and inhibiting Li dendritic growth. While the dataset is numerically small (N = 12), the electrolytes were selected to cover the vast majority of properties, spanning diverse solvation chemistries, concentrations, and solvation power of the solvent. Thus, we maintain that these groups (DE/WSE, HWSE, HCE/LHCE) represent distinct physicochemical regimes rather than arbitrary clusters. The representative solvation structures, distinct Li⁺-solvent and Li⁺-anion interactions, and concentration-dependent EDL properties all suggest that f effectively captures the competition between thermodynamics and kinetics governing Li dendrite growth. More importantly, this highlights the mechanistic link underlying the principle. However, extending this descriptor to other solvent-based electrolytes or complex formulations in the future will require further rigorous electrochemical validation and systematic system selection.

To elucidate the fractal characteristics of Li deposits, SANS was performed on Li deposits obtained from electrolytes A/D/F/J/K. Figure 3h presents SANS intensity profiles for Li deposits on Cu foils under a current density of 1.0 mA cm⁻² and an areal capacity of 6.0 mAh cm⁻². By fitting the SANS intensity curves within the Q range from 0.02 nm⁻¹ to 0.1 nm⁻¹ in Fig. 3i, it becomes clear that as CE_Li increases, the slope of the intensity profiles in this region decreases. This trend indicates the fractal character evolution in Li deposits from the pronounced surface fractal structure (characterized by the slope approaching −3) toward smoother and less fractal surfaces (approaching the slope of −4)⁴³. Further, phase-field simulations were performed to model Li plating. The nucleation boundary conditions were informed by the interfacial charge distribution states revealed by our CMD and ML results. Given the critical role of interfacial charge distribution uniformity in controlling Li nucleation sites, simulations were designed with two distinct nucleation scenarios: uniform versus random distributions. The uniform nucleation model represents high-f electrolytes, which exhibit homogeneous surface charge. While the random nucleation model represents low-f electrolytes characterized by localized charge accumulation. The phase-field simulations serve as the necessary bridge between the microscopic descriptor and the macroscopic dendrite growth, which allows us to verify how the charge dynamics governed by f translate into macroscopic dendrite morphologies. Li deposits were simulated with varying numbers of nucleation sites (N = 1, 3, 5, 7, 9, 11, 13, and 15), and the corresponding results after the simulation time of 80 s are presented in Supplementary Fig. 25. Figure 3j contrasts simulated Li plating for random and uniform scenarios with nucleation sites of 11. With increasing simulation time, dendrite growth gradually intensifies in the random nucleation case, underscoring the detrimental impact of uneven nucleation distributions. Similar comparative morphologies for scenarios with nucleation sites of 5, 7, and 11 are further detailed in Supplementary Fig. 26. To demonstrate the influence of nucleation uniformity on ECSA, the statistical analysis of the ECSA of Li dendrites at individual nucleation sites were conducted. Variances in dendritic ECSA among different nucleation sites were compared across different simulation parameters (number of nucleation sites, nucleation uniformity, and simulated time). As illustrated in Fig. 3k, the variance in ECSA under uniform conditions is substantially lower than under random conditions. Furthermore, as the number of nucleation sites increases in uniform conditions, this variance diminishes gradually, highlighting the critical importance of interfacial charge uniformity in achieving homogeneous Li plating. Additionally, synchrotron XCT was employed to validate the morphology and uniformity of Li deposits in different electrolytes. Figure 3l–n visually confirms that the Li deposit from electrolyte A exhibits a highly irregular morphology. In contrast, the Li deposit from WSE (electrolyte F) demonstrates notably improved homogeneity. Most significantly, the Li plating in LHCE exhibits enhanced uniformity, corroborating the relationship between interfacial charge redistribution dynamics and Li deposit morphology.

According to the definition of f, interfacial excess charge dynamics are governed by two principal aspects: the level of electrode excess charge and the rate of interfacial charge depletion. As illustrated in Fig. 4a, the vertical axis represents the Li⁺ concentration for various electrolytes. With increasing Li⁺ concentration, the primary contributor to interfacial excess charge gradually shifts from electrode excess charge toward cation screening effects. Meanwhile, the horizontal axis indicates the rate of interfacial charge depletion, which increases and eventually dominates as Li⁺-dipole interactions weaken. Consequently, electrolytes D to K can be clearly categorized into two distinct regimes based on Li⁺ concentration and Li⁺-dipole interactions for low-concentration groups (DE/WSE) and high-concentration groups (HCE, LHCE, and HWSE). In DE/WSE, interfacial excess charge dynamics shift from being controlled by electrode excess charge toward the depletion-rate dominated mechanism as Li⁺-dipole interactions weaken. Conversely, particularly in HCE, interfacial charge dynamics are primarily governed by cation screening. However, with the reduced Li⁺-dipole interactions present in LHCE and HWSE, enhanced interfacial charge depletion rates increasingly promote uniform Li plating. Revisiting the previously noted difference in CE_Li between electrolytes F and G, it becomes apparent that electrolyte F, characterized simultaneously by a higher interfacial charge depletion rate and a lower electrode excess charge, exhibits rapid charge redistribution dynamics for high CE_Li. Extending this trend of weakening Li⁺-dipole interactions further, we anticipate the existence of additional electrolyte regimes, such as localized HWSE (LHWSE) and enhanced WSE (eWSE), which would potentially offer greater interfacial charge uniformity and electrochemical performance.

Fig. 4. Mechanism of interfacial excess charge redistribution and performance of the 500 Wh kg⁻¹ lithium metal battery (LMB).

a Map illustrating how Li⁺ concentration and the Li⁺-dipole interaction jointly affect the interfacial charge redistribution dynamics. Differences in electrolyte concentrations and Li⁺-dipole interactions can lead to variations in the redistribution principles of interfacial excess charge, with predicted regions for localized high-concentration weak solvation electrolyte (LHWSE) and enhanced weak solvation electrolyte (eWSE) marked. Schematic demonstration of interfacial charge redistribution and related Li dendrite-growth mechanism for DE (b), WSE (c), and HCE/LHCE (d). Symbols: circles (Li⁺), ellipses (solvents), and diamonds (anions). e The average Coulombic efficiency for Li plating/stripping (CE_Li) for the proposed LHWSE. f Cycling performance of ~6.4 Ah Li| |NCM811 pouch cells with electrolyte A/I/K and LHWSE. The inset optical photograph shows cell dimensions.

In Fig. 3e–g, the relationship between interfacial excess charge dynamics and changes in ECSA during Li dendrite growth is quantitatively investigated, which establishes the influence of three critical parameters. To visually elucidate their impact on dendrite formation, schematic diagrams in Fig. 4b–d correlate interfacial charge redistribution behaviors with their respective Li plating morphologies. In Fig. 4b for DE, the interface initially accumulates significant excess charge. However, strong Li⁺-dipole interactions limit the rate of interfacial charge depletion, resulting in sluggish excess charge redistribution dynamics. This slow redistribution intensifies localized charge accumulation, triggering heterogeneous nucleation sites and promoting dendritic Li growth. In contrast, weakening the Li⁺-dipole interactions in WSE (Fig. 4c) not only decreases electrode surface excess charge but simultaneously accelerates the rate of interfacial charge depletion. This enhanced depletion dynamic facilitates rapid interfacial charge redistribution, resulting in a more uniform charge distribution across the entire electrode interface. Consequently, Li plating under these conditions exhibits less dendritic morphology. In LHCE and HWSE (Fig. 4d), densely arranged cations within the EDL impose uniform charge distribution through strong cation screening effects. Even though the rates of interfacial charge depletion may initially remain modest in typical HCE groups, the intrinsic uniformity provided by cation screening enhances homogeneous Li nucleation. Moreover, further reductions in Li⁺-dipole interactions in these HCEs enhance the rate of charge depletion. This synergy ultimately promotes planar Li plating, boosting the reversibility of Li plating/stripping. While our framework focuses on interfacial excess charge distribution, we acknowledge that this model simplifies the role of SEI. We have assessed the potential deviations that would arise if the SEI were integrated (Supplementary Note 6). We minimized the influence of SEI composition through electrolyte formulation control and the preconditioning protocol. However, this simplification is intentional because it isolates the fundamental competition between interfacial kinetics and thermodynamics governing dendrite growth. We posit that interfacial excess charge redistribution acts as the upstream governor of SEI stability. A high f value indicates the rapid dissipation of localized charge accumulation. This resulting uniformity in charge and ionic flux is the prerequisite for forming the uniform and stable SEI, thereby preventing the localized stress that leads to SEI rupture.

Based on the analysis illustrated in Fig. 4a, which differentiates electrolyte regions according to Li⁺-dipole interactions and Li⁺ concentration, we propose that emerging electrolyte regimes such as LHWSE and eWSE, characterized by elevated interfacial charge depletion rates and reduced electrode excess charge, will inherently exhibit improved CE_Li. Especially, regarding the LHWSE, we adapted its formulation directly from electrolyte K (LHCE) by replacing the DME with DEE, a solvent with weaker dissociation capability, and the volume of TTE is further increased to weaken the potential Li⁺-dipole interactions. In Fig. 4e, the resulting LHWSE achieved an average CE_Li of 99.55%. Driven by enhanced cation screening at higher Li⁺ concentrations and diminished Li⁺-dipole interactions induced by DEE and diluent, the electrode excess charge and exchange current density of LHWSE closely align with those of LHCE (Supplementary Fig. 27). SANS result reveals that Li deposits obtained from LHWSE exhibit notably smoother and less fractal morphologies (Supplementary Fig. 28). To validate the electrode excess charge redistribution ability of LHWSE at the device level, a comparative test was conducted, as illustrated in Fig. 4f. Equal masses of electrolyte A, I, K and LHWSE were injected to the ~6.4 Ah LMBs under low N/P ratio of 0.80 and lean electrolyte-to-capacity (E/C) ratio of 1.00 g Ah⁻¹. Specific details of the cell parameters and masses are provided in Supplementary Fig. 29 and Table 10, both cells can achieve specific energy beyond 500 Wh kg⁻¹. As illustrated, the LMB with electrolyte A, primarily governed by electrode surface excess charge, exhibited the poorest cycling stability at 0.1 C (1 C = 6500 mA) charging/discharging rate and stack pressure of 800 kPa, with a marked capacity decline after 20 cycles. The LMB using electrolyte I exhibited improved stability due to the higher lithium salt concentration and the lower solvent dissociation ability. In electrolyte K (ether-based LHCE), a significant capacity drop occurred after 50 cycles under the same conditions. Notably, the LMB with LHWSE consistently maintained specific energy above 500 Wh/kg throughout the first 60 cycles and retained approximately 97.3% (at a stack pressure of 800 kPa) of its capacity after 90 cycles. The cycling performance of two 1.2 Ah pouch cells (N/P ratio of 0.80) is also demonstrated with identical stability (Supplementary Fig. 30) at 0.1 C (1 C = 1200 mA) charging/discharging rate. The manufacturability and safety for LHWSE are discussed in Supplementary Note 7. The overlap between galvanostatic charge/discharge curves of LMB with LHWSE between the 10th and 50th cycles (Supplementary Fig. 31a) highlights the good electrochemical stability of the battery with LHWSE, indicating minimal internal polarization and stable cell-level ohmic resistance during cycling. To further investigate how interfacial excess charge redistribution influences Li nucleation uniformity, surface flatness analyses were performed on practical ~6.4 Ah LMB cells employing LHWSE and electrolyte A after 20 cycles at 0.1 C (Supplementary Fig. 31b, c). The cell employing LHWSE displays uniform and lower surface pressure distribution under identical external stack pressure, indicating minimal thickness expansion and highly uniform Li plating/stripping. Conversely, the cell with electrolyte A exhibits higher and more uneven surface pressure distributions. This result demonstrates the substantial efficiency of the proposed electrolyte strategy in enhancing cell lifetime under practical conditions.

Discussion

In summary, we demonstrate that interfacial excess charge redistribution dynamics, rather than static electrode excess charge, govern Li dendrite growth, providing a physicochemical framework for interpreting the hidden competition between charge accumulation and depletion. We introduced a quantitative descriptor, the interfacial excess charge distribution factor, which integrates the effects of electrode excess charge, exchange current density, and cation screening. By combining ML-assisted CMD with the proposed potentiostatic metric for ECSA evolution, the correlations between f and Li dendrite growth across diverse solvation regimes (DE, WSE, HCE, HWSE, LHCE) are established, which demonstrates that rapid interfacial excess charge redistribution capability significantly reduces ECSA growth rates, ultimately promoting higher CE_Li. We identified that interfacial excess charge redistribution is governed by the competitive interplay between two factors: (1) the magnitude of charge accumulation, influenced by electrode excess charge and cation screening; and (2) the depletion rate of interfacial excess charge, affected by Li⁺-dipole-anion interactions. Guided by these mechanistic insights, we rationally designed the LHWSE for optimized excess charge equilibration. Validating this strategy, a 6.4 Ah LMB with LHWSE delivers 515 Wh kg⁻¹ and demonstrates enhanced cycling stability compared to baseline formulations. Looking forward, independent validation of spatially resolved charge dynamics via high-resolution operando techniques, such as electrochemical atomic force microscopy or surface-sensitive spectroscopy, will be essential. Furthermore, extending this descriptor to other metal electrodes (e.g., Zn, Na, K) and complex multi-component electrolytes represents a critical next step to validate the universality of the charge redistribution principle beyond the Li-ether studied here. These findings highlight the importance of designing electrolytes with rapid interfacial excess charge redistribution capability by strategically considering Li⁺ concentration and Li⁺-dipole-anion interactions for advanced LMBs.

Methods

Materials

Lithium bis(fluorosulfonyl)azanide (LiFSI, purity ≥99.9%, battery grade), lithium hexafluorophosphate (LiPF₆, purity ≥ 99.99%, battery grade), lithium tetrafluoroborate (LiBF₄, purity ≥ 99.9%, battery grade), lithium difluoro(oxalato)borate (LiDFOB, purity ≥ 99%), 1,2-dimethoxyethane (DME, purity ≥ 99.9%, battery grade), 1,3-dioxolane (DOL, purity ≥ 99.9%, battery grade), 1,2-diethoxyethane (DEE, purity ≥ 99.8%, battery grade), 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropylether (TTE, purity ≥ 99.9%, battery grade), ethyl methyl carbonate (EMC, purity ≥ 99.99%, battery grade), dimethyl carbonate (DMC, purity ≥ 99.99%, battery grade), fluoroethylene carbonate (FEC, purity ≥ 99.9%, battery grade), diethyl carbonate (DEC, purity ≥ 99.99%, battery grade) were purchased from Suzhou DuoDuo Chem Co., Ltd. Cyclopentyl methyl ether (CPME, purity ≥ 99.9%, anhydrous) was purchased from Aladdin. Before using, all the solvents were treated with molecular sieves to remove any possible water. The detailed formulations for different electrolytes are listed in Supplementary Table 1. Li foil (450 μm) was used as received after purchase from Guangdong Canrd New Energy Co., Ltd. Cu foil (25 μm) for coin cell testing was purchased from Shenzhen Keijing Co., Ltd.

Characterizations

XPS tests were carried out on an ESCALAB 250Xi spectrometer (PHI Versaprobe 4) with Al Kα radiation. TOF-SIMS analysis (IONTOF-TOF-SIMS 5) was performed by a 10 kV Bi⁺ ion beam and 1 kV Cs⁺ sputtering over an area of 50 μm × 50 μm, and the current for the sputtering beam was steady and constant (60 nA). Prior to measurements, all Li symmetric coin cells were cycled 3 times for Li plating/stripping under a current density of 0.5 mA cm⁻² and an areal capacity of 0.5 mAh cm⁻². All battery tests were performed at 25 ± 1 °C. All samples were washed with DME, dried in an argon glovebox before testing, and then transferred into the XPS or TOF-SIMS analysis chamber by a vacuum transfer device to avoid direct contact with air and water.

Electrochemical methods

The Li | |Cu and Li | |Li coin cells were evaluated by using 2032-type coin cells with the Celgard 2325 separator with electrolyte of 50 μL. 450 μm Li foil and 25 μm Cu foil were employed as electrodes. E_plating was determined by using Li symmetric coin cells assembled with various electrolytes. All coin cells were assembled and crimped under identical conditions to ensure uniform internal pressure. The cells were cycled for 3 times for Li plating/stripping at a constant charge/discharge current density of 0.5 mA cm⁻² for 1 h using LANHE CT3004A. The average discharge plateau potential was defined as E_plating, including the contributions from concentration polarization and interfacial impedance. The E_PZC was measured using cells assembled with Li foil (450 μm) as the working electrode and Pt foil as the counter electrode; each electrolyte contained 0.02 mmol L⁻¹ ferrocene. EIS was conducted for these cells on a DH7006 electrochemical workstation using the single-frequency mode (1000 Hz) to monitor the variations of capacitance. A frequency of 1000 Hz was strategically selected for the single-frequency E_PZC to isolate the interfacial double-layer capacitance by excluding diffusion-limited and residual contributions, as previous works reported^44,45. The potential corresponding to the maximum capacitance in the impedance spectrum was designated as E_PZC. We acknowledge that determining E_PZC via single-frequency capacitance measurements (Supplementary Table 10) assumes the simplified electrical double-layer model, which carries the risk of conflating kinetic contributions if not carefully controlled in the non-Faradaic region and the applied potential range. The average CE_Li was determined using Li | |Cu coin cells and the modified protocol proposed by Zhang et al.⁴⁶. For each electrolyte, the test protocol proceeded as follows: a 10 h Li plating step was performed at 0.5 mA cm⁻², followed by a 10 min rest, the cell was then stripped to 0.5 V; then an additional 10 h Li plating step at 0.5 mA cm⁻² was carried out; then the cell underwent 10 cycles of 2 h stripping and 2 h plating at 0.5 mA cm⁻²; after 10 cycles of Li plating/stripping, the cell was stripped to 0.5 V at 0.5 mA cm⁻². Potentiostatic plating was performed on Li symmetric coin cells with various electrolytes; a potential bias of 25 mV was applied for 5400 s. After the plating step, the cells were immediately subjected to the potentiostatic electrochemical impedance spectroscopy (PEIS) step at the same applied potential over the frequency range from 100 kHz to 0.1 Hz (Supplementary Fig. 32). The current–time curves obtained during potentiostatic plating were used to evaluate Li dendrite growth. In the tests shown in Figs. 3b, c, 1.0 M LiPF₆ in DMC/EMC (vol. 1/1) served as the baseline electrolyte, and potentiostatic plating was conducted to analyze current evolution. Linear sweep voltammetry (LSV) was performed using a DH7006 electrochemical workstation at a scan rate of 1.0 mV s⁻¹ within the potential range of 3.0–6.0 V to test the electrochemical stability of electrolytes.

Pouch cell assembly and testing

15 pieces of double-coated NCM811 (areal loading of 26.0 mg cm⁻² on one side, active material ratio of 96.0% with 2.0% conductive material and 2.0% poly(vinylidene fluoride) binder) with tailor-made dimensions (6.2 cm × 7.2 cm) were employed as the positive electrodes. 20 μm Li (purchased from Tianjin China Energy Lithium Co., Ltd.) with tailor-made dimensions (6.4 cm × 7.4 cm) was employed as a negative electrode for a single positive electrode side. The cells were encapsulated in aluminum packing foil by hot-pressing, followed by electrolyte injection with a dosage of 6.5 g for the pouch cell. Pouch cells with certain stack pressures of 800 kPa were connected to the NEWARE system (China). The pouch cells were firstly charged at 0.05 C (1 C equals 6500 mA) to 4.3 V, and rested for 10 min, then the cells were discharged at 0.05 C to 2.8 V. For cycling testing, the rate for charge/discharge was 0.1 C (1 C equals 6500 mA) from 2.8 V to 4.3 V. Details for pouch cell parameters could be found in our previous work^47,48 and Supplementary Table 11. All battery electrochemical measurements were performed at a room temperature of 25 ± 1 °C.

In situ electrochemical ATR-FTIR

In situ electrochemical ATR-FTIR measurements were performed on a Thermo Fisher Scientific Nicolet iS50 spectrometer at 25 ± 1 °C under ambient laboratory conditions. Infrared signals were detected using a liquid-nitrogen-cooled broadband mercury cadmium-telluride (MCT) detector. For spectral acquisition, a ZnSe ATR crystal was used as the internal reflection element, and all the measurements were conducted at a fixed incidence angle of 60°, covering a range of 2000–650 cm⁻¹⁴⁹. Based on the evanescent-field penetration model and the refractive indices of the ZnSe crystal and the liquid electrolyte, the effective probing depth of the present ATR setup was estimated to be approximately ~1 μm. Such a detection depth enables the infrared response to reflect chemical variations occurring both near the electrode/electrolyte interface and in the adjacent electrolyte region. For in situ measurements, a home-built two-electrode electrochemical cell was assembled inside an Ar-filled glovebox, where the concentrations of H₂O and O₂ were both maintained below 0.5 ppm. In the cell, a Cu mesh (150 μm aperture size and 9 μm thickness) was pressed directly against the ZnSe window and served as the working electrode, whereas Li foil (450 μm) acted as the counter and reference electrode. A Whatman glass-fiber separator was placed between the two electrodes, followed by the introduction of 100 μL of electrolyte into the cell. To minimize spectral interference from atmospheric species during data collection, the optical chamber was continuously protected by a purified N₂ flow provided through the gas-cleaning module integrated with the FTIR system. To monitor the desolvation process at the Cu mesh surface, in situ ATR-FTIR spectra were collected during chronopotentiometry measurements conducted at a current density of 2.0 mA cm⁻².

Small-angle neutron scattering (SANS)

SANS measurements were conducted at the BL14 beamline of the China Spallation Neutron Source (CSNS). During the experiments, neutrons with wavelengths spanning 2.2–6.7 Å were used, and the instrument was operated with a collimation length of 12.75 m. To avoid air exposure, all sample handling and sealing procedures were performed inside an argon-filled glovebox. The prepared samples were enclosed with Kapton tape prior to measurement. For background correction, a bare Cu foil (25 μm) together with blank Kapton tape was measured separately as the reference background. The scattering data were then reduced using the direct-beam-based procedure⁵⁰, including solid-angle correction, direct-beam calibration, subtraction of the background scattering contribution, and transmission correction.

Synchrotron X-ray computed tomography (XCT)

Synchrotron XCT experiments were carried out at beamline 1.2W (1.2 GeV with a beam current of 150 mA) of the Synchrotron Light Research Institute in Thailand. The samples were fixed and sealed in a plastic tube using glue before mounting on the stage. During each tomography measurement, projection images were collected over a rotation range of 180° with an angular interval of 0.1°. X-ray images were recorded using a detection system composed of an X-ray microscope (Optique Peter, France), a YAG:Ce scintillator (100 μm-thick, Crytur, Czech Republic), and an sCMOS detector (Andor NEO 5.5, 2560 × 2160 pixels, 16 bit), yielding a final effective pixel size of 0.72 μm. The acquired projection data were subsequently subjected to background correction, pre-processing, and tomographic reconstruction using Octopus Reconstruction software (TESCAN, Gent, Belgium). The reconstructed datasets were finally visualized and rendered into three-dimensional images using Drishti software.

Molecular dynamics (MD) simulations

The Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS)⁵¹ package was employed to conduct all molecular dynamics (MD) simulations. To maintain temperature and pressure control, we employed the Nosé−Hoover thermostat and Anderson barostat⁵² method, incorporating relaxation constants of 100 fs and 1000 fs, respectively. Simulations were performed under both the canonical (NVT) and isothermal-isobaric (NPT) ensembles as required. Lorentz-–Berthelot (LB) mixing rule was applied to determine cross-interaction parameters⁵³, while the Long-range Coulombic interactions was computed via particle-particle-particle-mesh (PPPM) algorithm⁵⁴. The simulations were implemented in the NVT ensemble at 300 K for 5 ns. The time step for all simulations was set to 1 fs. The lattice constants of the supercells used in the simulations were 24.6 Å × 24.6 Å × 100.0 Å. The forcefield parameters for LiFSI, DME, DEE, and Li metal electrode (Li atoms) were adopted^55–58. The atomic charges of configurations were predicted by the charge module of QMTP, which was implemented in the EASYMLFF package⁵⁹. All the initial configurations of interface systems were constructed by Packmol⁶⁰. All configurations (Supplementary Data 1) were visualized using OVITO⁶¹ and VESTA software⁶².

Machine learning model

To achieve accurate prediction of atomic charge for large-scale systems, the QMTP model for the interfaces of Li metal electrode (Li atoms) with LiFSI/DME electrolyte and LiFSI/DEE electrolyte was trained³⁴. The QMTP model includes two modules, the charge module and the energy/force module. In this work, we only used the charge module. The model parameter of ${\mathrm{lev}}_{\max }^{\mathrm{Q}}$ was set as 12, and the parameter of ${\mathrm{r}}_{\mathrm{c}}^{\mathrm{Q}}$ was chosen to be 5.0 for atomic charge fitting. The training datasets were derived from AIMD simulations, sampling, and an active learning sampling strategy⁶³. More details about the machine learning model can be found in the open-access Zenodo repository⁶⁴ and the Supplementary Note 1/Table 4.

Phase-field simulations

The phase-field simulations were performed using COMSOL Multiphysics 6.2, with simulated dimensions of 6 × 6 µm². Dirichlet boundary conditions were applied to the Nernst-Planck equation and the current continuity equation, while a zero-flux boundary condition was imposed on the phase-field variable (ξ)⁶⁵. The boundary conditions set the Li-ion concentration to 1.0 mol L⁻¹ in the liquid electrolyte and 0.0 mol L⁻¹ in the LMA. The electric potential (ϕ) was fixed at 0.0 V in the liquid electrolyte and −0.2 V in the LMA. Two nucleation methods were employed, namely random nucleation and uniform nucleation. In the random nucleation model, semi-circular random noise was introduced on the surface of the LMA to serve as the nucleation sites. In the uniform nucleation, nucleation sites were uniformly distributed across the surface. Different numbers of nucleation sites were assigned to enable a comparative analysis of their effects on dendrite growth.

The calculation of interfacial charge distribution factors (f)

The f is defined as ${\mathrm{J}}_0$ normalized by ${\mathrm{\sigma }}_{\mathrm{net}}$. As discussed by Filhol²⁹ based on the Lippmann equation⁶⁶, electro-capillary curves can be approximated at second order by the following Expression 5, and the surface charge density ${\mathrm{\sigma }}_{\mathrm{surface}}$ is given by the Lippmann equation from Expression 6.

$$\gamma \left(\mathrm{E}\right)=-\frac{{\mathrm{C}}_{\mathrm{dl}}}{2}{\left(\mathrm{E}-{\mathrm{E}}_{\mathrm{PZC}}\right)}^2+{\gamma }_{\mathrm{PZC}}$$ 5

$${\mathrm{\sigma }}_{\mathrm{surface}}=-\frac{\partial \gamma }{\partial \mathrm{E}}$$ 6

where γ(E) and ${\gamma }_{\mathrm{PZC}}$ stand for the surface tension at a given potential (E) and the potential of zero charge (E_PZC), and ${\mathrm{C}}_{\mathrm{dl}}$ refers to the electric double-layer capacitance. E_PZC represents the thermodynamic equilibrium state where the electrode surface carries no net charge. However, the Li plating reaction occurs at a given potential (E_plating), which is the potential required to drive the reduction reaction of Li⁺. Consequently, to facilitate Li plating, the electrode potential must be polarized to E_plating that is much more negative than E_PZC. This deviation from the neutral state generates the negative electrode excess charge. According to Expression 5–6, the ${\mathrm{\sigma }}_{\mathrm{surface}}$ at the Li plating conditions can be calculated by using Expression 7.

$${\mathrm{\sigma }}_{\mathrm{surface}}=\left({\mathrm{E}}_{\mathrm{PZC}}-{\mathrm{E}}_{\mathrm{plating}}\right)*{\mathrm{C}}_{\mathrm{dl}}$$ 7

The power-law relationship is chosen to describe the non-linear response of interfacial charge accumulation to bulk concentration. Under the Debye–Hückel approximation, the differential capacitance (${\mathrm{C}}_{\mathrm{diff}}$) of the diffuse layer is directly proportional to the inverse Debye length ($\mathrm{\kappa }$) in Expression 8.

$${\mathrm{C}}_{\mathrm{diff}}=\mathrm{ϵ}{\mathrm{ϵ}}_0\mathrm{\kappa }$$ 8

The characteristic thickness of the screening layer, known as the Debye length (${\lambda }_{\mathrm{D}}$), is determined by the bulk ionic concentration (${\mathrm{c}}_0$) and solvent properties, its reciprocal is defined as Expression 9:

$${\lambda }_{\mathrm{D}}^{-1}=\mathrm{\kappa }=\sqrt{\frac{2{\left(\mathrm{ze}\right)}^2{\mathrm{N}}_{\mathrm{A}}{\mathrm{c}}_0}{ϵ{\mathrm{ϵ}}_0{\mathrm{k}}_{\mathrm{B}}\mathrm{T}}}$$ 9

where $ϵ$ is the dielectric constant, ${\mathrm{ϵ}}_0$ is the vacuum permittivity, ${\mathrm{k}}_{\mathrm{B}}$ is the Boltzmann constant, $\mathrm{T}$ is the temperature, $\mathrm{z}$ is the valence, $\mathrm{e}$ is the elementary charge, and ${\mathrm{N}}_{\mathrm{A}}$ is the Avogadro’s constant. The screening charge density is the total countercharge stored in the diffuse layer, related to the effective potential drop (${\mathrm{\Delta }\mathrm{\varphi }}_{\mathrm{eff}}$) across the layer, as described in Expression 10:

$${\mathrm{\sigma }}_{\mathrm{screen}}=ϵ{\mathrm{ϵ}}_0\mathrm{\kappa }{\mathrm{\Delta }\mathrm{\varphi }}_{\mathrm{eff}}$$ 10

Substituting $\mathrm{\kappa }$ into Expression 10 then produces the following Expression 11:

$${\mathrm{\sigma }}_{\mathrm{screen}}=\left[{\mathrm{\Delta }\mathrm{\varphi }}_{\mathrm{eff}}{\left(\frac{2{\left(\mathrm{ze}\right)}^2{\mathrm{N}}_{\mathrm{A}}\mathrm{ϵ}{\mathrm{ϵ}}_0}{{\mathrm{k}}_{\mathrm{B}}\mathrm{T}}\right)}^{0.5}\right]\cdot {{\mathrm{c}}_0}^{0.5}=\mathrm{\Upsilon }\cdot {{\mathrm{c}}_0}^{0.5}$$ 11

The pre-factor $\mathrm{\Upsilon }$ is a physically constant; however, the determination of the specific dielectric environment ($ϵ{\mathrm{\Delta }\mathrm{\varphi }}_{\mathrm{eff}}$) for the electrolyte is important in future works. Based on the relationships described in Expression 5–11, the overall relationship of the interfacial charge distribution factor can be generalized in Expression 2. Furthermore, we rationalize the use of ${\mathrm{c}}_0$ in the screening term based on classical electric double-layer theory. While the local concentration at the interface certainly deviates from the bulk, the characteristic ${\lambda }_{\mathrm{D}}$ and the diffuse layer capacitance are mathematically determined by the bulk ionic strength. The interfacial gradients are merely the thermodynamic response to this baseline. Therefore, utilizing ${\mathrm{c}}_0$ as the scaling variable correctly captures the driving force for cation accumulation. Our treatment of cation screening relies on the application of classical Gouy–Chapman and Grahame theories. The $\mathrm{\Upsilon }$ can be calibrated against the baseline electrolyte. While this approach can capture the dominant concentration dependence, it serves as a simplified representation of the complex ion-ion correlations and adsorption effects present in concentrated multi-component electrolytes.

Measurement of exchange current density

To capture interfacial kinetics under conditions consistent with dendrite growth, we employed a combined chronoamperometry (CA, potentiostatic plating) and PEIS method (Supplementary Fig. 32). Prior to measurement, all Li symmetric coin cells were cycled for 3 times of Li plating/stripping under 0.5 mA cm⁻² and 0.5 mAh cm⁻². This preconditioning procedure was intended to ensure the removal of the native oxide layer and the establishment of the stable interphase. XPS and TOF-SIMS results (Supplementary Figs. 2 and 3) confirm that the electrode surface reaches a chemically steady state following this protocol. Following preconditioning, a potentiostatic plating step was performed at the potential bias of 25 mV (η_total) for 5400 s. Crucially, to preserve the polarization state, PEIS was performed immediately at the same bias potential over a frequency range from 100 kHz to 0.1 Hz without a resting period. Since the applied potential during plating includes ohmic drops from the bulk electrolyte and interphase, it doesn’t represent the driving force for charge transfer. To correct for this, PEIS data were deconvoluted using the distribution of relaxation times (DRT) method. This allowed for the separation of bulk resistance (${\mathrm{R}}_{\mathrm{s}}$), interphase resistance (${\mathrm{R}}_{\mathrm{SEI}}$), and charge transfer resistance (${\mathrm{R}}_{\mathrm{ct}}$). The effective overpotential (η_ct) driving the electrochemical reaction was calculated by partitioning the total applied potential according to Expression 12.

$${\mathrm{\eta }}_{\mathrm{ct}}={\mathrm{\eta }}_{\mathrm{total}}*\frac{{\mathrm{R}}_{\mathrm{ct}}-{\mathrm{R}}_{\mathrm{s}}}{{\mathrm{R}}_{\mathrm{SEI}}-{\mathrm{R}}_{\mathrm{s}}}$$ 12

where ${\mathrm{\eta }}_{\mathrm{total}}$ is the overpotential applied over the cells, ${\mathrm{R}}_{\mathrm{s}}$, ${\mathrm{R}}_{\mathrm{ct}}$, and ${\mathrm{R}}_{\mathrm{SEI}}$ are obtained from PEIS results. J₀ was then derived using the linear approximation of the Butler-Volmer equation as Expression 13, valid for small overpotentials:

$${\mathrm{J}}_0={\mathrm{\eta }}_{\mathrm{total}}*\frac{\mathrm{iRT}}{{\mathrm{\eta }}_{\mathrm{ct}}\mathrm{F}*\mathrm{A}}$$ 13

where i is the current density at the end of the potentiostatic step, $\mathrm{F}$ is the Faraday constant, $\mathrm{R}$ is the universal gas constant, and $\mathrm{T}$ is the room temperature. To verify the robustness of this method, measurements were repeated for independent cells per electrolyte. The calculated standard deviations and confidence intervals confirm the high reproducibility of this methodology. The exchange current density is highly sensitive to the electrode’s cycling history and preconditioning protocols. While we mitigated these via identical preconditioning and DRT decoupling, we recognize that these parameters represent effective values under our specific experimental conditions.

Calculation of ECSA growth rate via potentiostatic transients

According to the Cottrell equation (Eq. 14), the time-dependent current density ($\mathrm{i}$) in the diffusion-controlled and diffusion-limited regime is expressed as:

$$\mathrm{i}=\frac{\mathrm{nF}{\mathrm{D}}^{1/2}{\mathrm{c}}_0\mathrm{A}}{\sqrt{\mathrm{\pi }}\mathrm{t}}$$ 14

where $\mathrm{n}$ is the number of electrons transferred, $\mathrm{F}$ is the Faraday constant, $\mathrm{D}$ is the diffusion coefficient of electrolyte, ${\mathrm{c}}_0$ is the bulk concentration of electrolyte, $\mathrm{A}$ is the ECSA of electrode, and $\mathrm{t}$ is time. Under ideal potentiostatic electroplating conditions with the constant ECSA, the current in the diffusion-limited regime is inversely proportional to $\sqrt{\mathrm{t}}$. To evaluate the variation in ECSA, the current–time curve in the diffusion-limited region is analyzed (Supplementary Fig. 23). We define the inflection point (onset of diffusion control) as (${\mathrm{t}}_1$, ${\mathrm{i}}_{\mathrm{t}1}$), and the endpoint as (${\mathrm{t}}_3$, ${\mathrm{i}}_{\mathrm{t}3}$) in Fig. 3a. We assume ${\mathrm{k}}_1$ represents the rate of ECSA variation, corresponding to dendrite growth rate. We assume the ECSA at ${\mathrm{t}}_1$ is ${\mathrm{A}}_1$, and that ECSA evolves over time (${\mathrm{A}}_{\mathrm{n}}$) according to the linear growth function defined in Expression 15:

$${\mathrm{A}}_{\mathrm{n}}={\mathrm{A}}_1*\left[1+{\mathrm{k}}_1*\left({\mathrm{t}}_{\mathrm{n}}-{\mathrm{t}}_1\right)\right]$$ 15

Substituting Expression 15 into the Cottrell equation (Eq. 14) gives the instantaneous current (${\mathrm{i}}_{\mathrm{tn}}$) at time ${\mathrm{t}}_{\mathrm{n}}$ as Expression 16:

$${\mathrm{i}}_{\mathrm{tn}}=\frac{\mathrm{nF}}{\sqrt{\pi }}*{\mathrm{D}}_1^{\frac{1}{2}}*{\mathrm{c}}_1*\frac{{\mathrm{A}}_1}{\sqrt{{\mathrm{t}}_1}}*\left[1+{\mathrm{k}}_1*\left({\mathrm{t}}_{\mathrm{n}}-{\mathrm{t}}_1\right)\right]$$ 16

When extending this analysis to a second electrolyte with inflection point (${\mathrm{t}}_2$, ${\mathrm{i}}_{\mathrm{t}2}$) and endpoint (${\mathrm{t}}_4$, ${\mathrm{i}}_{\mathrm{t}4}$) in Fig. 3a, the direct comparisons of ECSA are challenging due to the differences in transport properties ($\mathrm{D}$, ${\mathrm{c}}_0$) and initial ECSA among different electrolytes. To address this, as discussed in Supplementary Note 3, we employ a self-normalization strategy. We observe that the transport properties ($\mathrm{D}$, ${\mathrm{c}}_0$) and initial ECSA for each electrolyte can be mathematically substituted using the instantaneous current and time values at the respective initial inflection points, as detailed in Expression 17. Specifically, the ratio of transport parameters and initial ECSAs between two electrolytes can be expressed solely through the relationship of their initial current–time coordinates.

$$\frac{{\mathrm{i}}_{\mathrm{t}1}*\sqrt{{\mathrm{t}}_1}}{{\mathrm{i}}_{\mathrm{t}2}*\sqrt{{\mathrm{t}}_2}}=\frac{{\mathrm{D}}_1^{\frac{1}{2}}*{\mathrm{c}}_1*{\mathrm{A}}_1}{{\mathrm{D}}_2^{\frac{1}{2}}*{\mathrm{c}}_2*{\mathrm{A}}_2}$$ 17

This ultimate goal is related to the slope of the rising current to the ECSA growth. By applying the above Expressions, Expression 18 is derived.

$$\frac{\mathrm{\Delta }{\mathrm{i}}_{\mathrm{t}4-\mathrm{t}2}}{\mathrm{\Delta }{\mathrm{i}}_{\mathrm{t}3-\mathrm{t}1}}*\frac{{\mathrm{i}}_{\mathrm{t}1}*\sqrt{{\mathrm{t}}_4}}{{\mathrm{i}}_{\mathrm{t}2}*\sqrt{{\mathrm{t}}_3}}=\frac{\left[{\mathrm{k}}_2*\mathrm{\Delta }{\mathrm{t}}_{4-2}*\sqrt{{\mathrm{t}}_4}+\sqrt{{\mathrm{t}}_2}-\sqrt{{\mathrm{t}}_4}\right]}{\left[{\mathrm{k}}_1*\mathrm{\Delta }{\mathrm{t}}_{3-1}*\sqrt{{\mathrm{t}}_3}+\sqrt{{\mathrm{t}}_1}-\sqrt{{\mathrm{t}}_3}\right]}$$ 18

Simplifying Expression 18 yields the ratio of growth rate (k₂/k₁), as shown in Expression 19.

$$\frac{{\mathrm{k}}_2}{{\mathrm{k}}_1}=\frac{\mathrm{Slope}2*{\mathrm{i}}_{\mathrm{t}1}}{\mathrm{Slope}1*{\mathrm{i}}_{\mathrm{t}2}}*\left(1+\frac{\sqrt{{\mathrm{t}}_1}-\sqrt{{\mathrm{t}}_3}}{\mathrm{\Delta }{\mathrm{t}}_{3-1}*\sqrt{{\mathrm{t}}_3}}\right)-\frac{\sqrt{{\mathrm{t}}_2}-\sqrt{{\mathrm{t}}_4}}{\mathrm{\Delta }{\mathrm{t}}_{4-2}*\sqrt{{\mathrm{t}}_4}}$$ 19

where Slope1 and Slope2 represent the slopes of the linear current rise in the diffusion-limited region for each electrolyte. Mathematically, this derivation ensures that the calculated ECSA growth is independent of the electrolyte’s bulk diffusivity. By canceling out the transport terms, the resulting metric serves as the indicator of interfacial morphological instability. For the determination of ECSA growth rate via potentiostatic plating, the appropriate selection of polarization voltages and time windows is necessary to ensure the cell enters the diffusion-limited region and follows the Cottrell decay. This method is most accurate for electrolytes where the Coulombic efficiency is reasonably high (> 95%). We also note that this method may not be directly applicable to conventional porous current collectors due to variations in their unstable specific surface area. Additionally, for metallic electrodes with artificial interphases, the decoupling of impedance contributions from the artificial layer versus dendritic surface area expansion requires further validation.

Author contributions

W.D., X.L., J.Z. and Y.T. conceived the concept and the project. W.D. and X.L. performed the experiments and data analysis. G.L., C.F., S.C. and J.L. performed theoretical calculations. W.D. and Z.W. assembled and tested the pouch cells. W.D., X.L. and P.P. performed the XCT and data analysis. X.L., T.Z, H.C. and W.D. performed the SANS and data analysis. Y.T. supervised the work. W.D., X.L., G.L., B.Q., X.Z., J.Z. and Y.T. wrote and revised the manuscript. All authors discussed the results and reviewed the manuscript.

Peer review

Peer review information

Nature Communications thanks Hansen Wang and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The datasets analyzed and generated during the current study are included in the paper and its supplementary information. Source data are provided with this paper.

Code availability

The training datasets and the universal machine learning potential developed in this study have been deposited in the open-access Zenodo repository. The digital object identifier (DOI) for the publicly accessible data is as follows: 10.5281/zenodo.19248315.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-72472-y.