Hidden subsurface molecular bubbles in graphite anodes for LIBs

Abstract

The interplay between solvent co-intercalation, solid–electrolyte interface (SEI) formation, and gas evolution at the graphite anode–electrolyte interface plays a critical role in battery performance; yet, it remains poorly understood at the nanoscale. In this study, we introduce ultrasound-based operando atomic force microscopy (AFM), which breaks the spatial-resolution limitation of ultrasound-based techniques, to visualize the dynamics of solvent co-intercalation, SEI formation, and subsurface gas evolution in graphite anodes for lithium-ion batteries. Remarkably, we observe that gas evolution leads to the formation of “subsurface molecular bubbles”—gaseous pockets trapped between graphite layers—that compromise interfacial stability during battery formation cycles. AFM and density functional theory calculation results revealed that these subsurface molecular bubbles are primarily induced by the co-intercalation and decomposition of Li⁺(EC)₄ solvation complexes. We also found that the solvent co-intercalation and interlayer decomposition effects can be fully suppressed by incorporating a low-permittivity, non-solvating diluent solvent (fluoride benzene) through optimizing the de-solvation energy and the interfacial molecular architectures. By applying this optimized electrolyte in both graphite/Li half-cells and lithium cobalt oxide (LCO)/graphite full-cells, we achieve stable cycling with negligible molecular bubble formation, compact SEI growth, and high coulombic efficiency (>93%) during high-rate (0.5C) battery formation.

---

Broader context

The drive for cost-effective lithium-ion battery (LIB) production has intensified efforts to shorten formation cycles; yet, accelerated protocols often degrade electrode–electrolyte interfaces, impairing performance. While solid–electrolyte interphase (SEI) formation is well-studied, atomic-scale interfacial dynamics during early cycling remain poorly understood, hindering rational process optimization. Here, operando atomic force microscopy with near-field ultrasonic excitation uncovers hidden subsurface molecular bubbles in graphite anodes, formed by trapped decomposition gases between graphite carbon layers during Li⁺(EC)₄ co-intercalation. These bubbles mechanically destabilize the SEI, increasing irreversible capacity losses. Crucially, we demonstrate that non-solvating diluent solvents suppress bubble formation by tuning solvation structures and de-solvation energetics, enabling stable full-cell operation under rapid formation. By linking molecular-scale interface evolution to macroscopic performance, this work provides both fundamental insights into a previously overlooked degradation pathway and actionable strategies for designing efficient formation protocols—critical advances toward scalable, sustainable energy storage systems.

1. Introduction

The conventional formation process of commercial lithium/sodium ion batteries (LIBs/SIBs) requires low C-rate charge/discharge cycles to form a homogeneous, chemically and mechanically robust solid-state interphase (SEI).^1,2 During the formation cycles, the electrolyte decomposition on the graphite anode generates not only a protective SEI but also detrimental gas products that can damage the micro-structural integrity of the anode and SEI.^3,4 To reduce the time for the overall formation process, a deeper understanding of interfacial structure–property correlations between solid SEI components and gaseous products is urgently required.

The formation of the SEI and gas evolution in graphite anodes are closely interrelated interfacial phenomena. The SEI is a result of electrolyte decomposition, which forms a passivating film that separates the reactive electrode from the electrolyte.⁵ Gas evolution, on the other hand, is often associated with the decomposition of the electrolyte and the subsequent reactions occurring at the graphite–electrolyte interface.^3,6 Interestingly, the SEI formation often involves the generation of gases (carbon dioxide and hydrocarbons^7–9), and these gases, in turn, can affect the formation and stability of the SEI, causing uneven ion transportation at interfaces³ and ultimately leading to diminished shelf life and battery lifetime. So far, various measures, such as electrolyte additives, surface modification, and electrode engineering, are being explored to enhance the stability of the SEI and mitigate gas evolution in graphite anodes.^6,10,11 However, when it comes to interfacial structural degradation, gas evolution on the electrode–electrolyte interface at the nanoscale is often overlooked.¹² For example, the lost initial coulombic efficiency (ICE) of the graphite anode is often attributed to the SEI formation, but much less attention has been paid to anode structural damage due to gas evolution on the graphite–electrolyte interface.

It is intuitively reasonable to assume that the gases generated at the edge planes of graphite can diffuse into both the graphite interlayers and liquid electrolyte. The gas released into the electrolyte can be sandwiched in hydration layers in the graphite–liquid interface^13,14 or can form macroscale bubbles inside the battery pouches, which has been studied extensively by chromatographic mass spectrometry,¹⁵ X-ray tomography,¹⁶ neutron radiography¹⁷ and ultrasonic non-destructive testing.^18,19 In contrast, the gases that diffuse into (or are generated inside) the graphite interlayers may be trapped inside the solid lattice²⁰ and may damage the graphitic lattice. The trapped gas is rarely explored and its effects are far from being understood due to the lack of nanoscale interfacial characterization techniques. Although traditional in-situ/operando electrochemical atomic force microscopy (EC-AFM) has shown promising potential to reveal the nanoscale morphological interfacial structure and property evolution,^11,21 it still lacks subsurface characterization capability that can probe the trapped gas. For example, the nano-bumps observed on the carbon anode surface were previously interpreted as the nanoblisters²² filled by co-intercalated liquid solvent between the carbon layers, but it was unclear the whether there are gas products inside the bumps. Another example is oxygen redox in the cathode lattice,²³ which has raised more and more research interest; yet, it is still challenging to detect the trapped gas molecules experimentally.^24–26 The large difference in the acoustic impedance of gas vs. liquid and solid phases suggests deploying ultrasound-based techniques to study the buried gas evolution behaviour on battery interfaces.

Thanks to the non-invasive and non-destructive characteristics of ultrasound, ultrasound-based techniques enable in situ/operando monitoring of gas evolution in the internal structures of a battery.^27–29 These ultrasound-based techniques exhibit high sensitivity to changes in the physical and chemical properties of materials inside the battery package, allowing the detection of small variations and transient gas evolution behaviour, as well as the identification of potential issues or abnormalities within the battery. However, the above-mentioned ultrasound-based characterization techniques, such as the ultrasound-based 2D imaging method,²⁷ merely allow the millimetre-scale localization of gas evolution within the pouch cell.²⁸ This is because the spatial-resolution (10^4–6 nm) of the ultrasound-based technique is limited by the wavelength of ultrasound in the sample. To study the nanoscale sized graphite–electrolyte interfaces during gas evolution and to facilitate the comprehensive investigation of the origins of macroscale gas evolution processes, it is necessary to bridge the gap between these two by a non-invasive technology that cannot be provided by standard ultrasound techniques. Electrochemical ultrasonic force microscopy³⁰ (EC-UFM) uses ultrasound as a tool to probe local nanomechanical properties via the highly localized AFM tip contact, achieving resolution dictated by the tip size, not the sound waves. Therefore, EC-UFM is a powerful tool to study the buried gas evolution behaviour, in situ and operando, with nanoscale resolution.^31,32

In this work, we introduced a novel ultrasound-based operando atomic force microscopy technique (Notes S1 and S2, ESI†) to study the complex interrelations of SEI formation and gas evolution on the graphite–electrolyte interface at the nanoscale. This directly imaged microscopic subsurface “molecular bubbles” trapped inside graphene interlayers, revealing the previously largely ignored factor of interfacial structural degradation during battery formation. We find that it is solvent co-intercalation and decomposition between the carbon interlayers that are the main causes of subsurface molecular bubble formation. To complement UFM observations, density functional theory (DFT) calculations predict that Li⁺(EC)₄ is the most likely solvation complex that participates in both co-intercalation and decomposition. To inhibit this interfacial degradation, we tune the electrolyte component by adjusting the dipole–dipole interactions using a low-dielectric constant diluent and then thoroughly investigate in-depth intermolecular interactions and interfacial electrolyte structures of our optimized electrolytes using molecular dynamics (MD) simulation and force distance spectroscopy. The optimized electrolyte enables fundamental suppression of the formation of subsurface molecular bubbles and the SEI, with the intact graphite–electrolyte interface improving the cycle stability and coulombic efficiency of battery cells under an elevated current density (0.5C) during the battery formation process.

1.1. Macroscale and nanoscale graphite interfacial degradation during the formation process

Fig. 1a shows the galvanostatic charge/discharge curves of a LiCoO₂/graphite full-cell and the LiCoO₂ cathode using a three-electrode cell at formation current densities of 0.05C and 0.5C. At a high formation rate (0.5C), fluctuations in the full-cell charge/discharge curves are observed in addition to the decreased reversible capacity expected for the high formation rate. By contrast, the cathode voltage curve stays smooth and stable, indicating that the formation rate of the full cell is limited by the graphite anode.³³ According to optical microscopy observations (Fig. 1b), the decrease in reversible capacity is partially attributed to the accelerated electrolyte decomposition and limitations in ionic diffusion. In this cell configuration, bubble formation due to electrolyte decomposition on the electrode surface can be clearly observed from the glass window at the top of the cell as shown in Fig. 1b. These bubbles accumulate at the electrode–glass interface (Fig. S1 and Video S1, ESI†) and disrupt the electrical contact, causing fluctuations in the cell's internal resistance. The gaseous products accumulated on the graphite–electrolyte interface separate the lithium storage sites and electrolyte, leaving a large proportion of graphite particles not fully lithiated into LiC₆ according to operando Raman spectroscopy (Fig. S2a, ESI†). Therefore, the macroscale gas evolution and accumulation on the graphite–electrolyte interface was found to be the “visible” cause of incomplete lithiation and the consequent capacity loss during the battery formation.

Fig. 1 Macroscale and nanoscale interfacial structural degradation of the graphite anode during the initial formation in 1 M LiPF₆ in EC : DEC = 1 : 1 electrolyte. (a) The initial charge and discharge curves of the LiCoO₂/graphite cell in a three-electrode cell (ECC-Opto-Std test cell) during the formation cycles. The blue solid lines are the full-cell voltage; the red dashed lines are the LiCoO₂ cathode voltage. (b) Schematic diagram of an electrochemical cell for operando optical microscopy and the optical image of a gas bubble in a cell during the 1st formation cycle at 0.5C (Video S1, ESI†). (c) Schematic diagram of a graphite anode measured by conventional electrochemical atomic force microscopy (EC-AFM) during the formation charging (∼0.4C). The galvanostatic charge curves are shown beside the diagram. (d) Surface topography changes of the graphite anode observed by operando EC-AFM during the lithiation at different charge states No. 1–4 as denoted in Fig. 1c. See the video of operando scanning observation (Video S2, ESI†). (e) Schematic diagram of the formation of “nano-bumps” and SEIs formed on a graphite anode surface. (f) The statistical results of the maximum height vs. radius, R, ratio of the nano-bumps. The effective radius R determined as , and for circular, trapezoidal and triangular nano-bumps, respectively, where S is the area of the bump. The shape-coefficient c = 0.7 was used for the fitting of the adhesion energy (γ) for all types of nano-bumps, see eqn (1) for details. (g) and (h) The mechanical modulus peak-force images of the nano-bumps at charge stages No. 2 and No. 3 in Fig. 1d.

Operando EC-AFM (Fig. 1c) was used to study the nanoscale “invisible” interfacial degradation processes on the graphite anode surface further. As shown in Fig. 1d, the evolution of the surface morphology of the basal planes on a graphite particle surface is observed by EC-AFM at open circuit potential (OCP) and different charging voltages (No. 1–4). Initially (at No. 1 OCP), the basal planes of the graphite particle surface are smooth and clean (as confirmed by the SEM image in Fig. S2b, ESI†) due to the atomic flat characteristics of the carbon (001) plane. Surprisingly, we observed the nucleation and growth of multiple “nano-bumps” during galvanostatic charging (see Video S2, ESI†). The first nano-bump appears on the flat graphite surface at voltage region No. 1, after which the size of the nano-bump increases with the lithium-ion intercalating into the graphite. At the end of the charge, in addition to the nano-bumps formed on the anode surface, we observed dispersed SEI nanoparticles covering the graphite surface (see the AFM image at voltage region No. 4 in Fig. S2c and S2d, ESI†). These nano-bumps are a result of local delamination (as outlined in Fig. 1e), similar to the previously reported gas bubbles or liquid blisters trapped between the layers of 2D materials.^34,35Fig. 1f shows the statistical value of the maximum height vs. reduced radius ratio (h_max/R) of the nano-bumps for different nano-bump shapes. The aspect ratio (h_max/R) of the nano-bumps is related to the total adhesion energies γ by³⁶

where Y is Young's modulus of graphene, c is the shape-coefficient and γ is the total adhesion energy linked to the van der Waals force between graphene and the substrate,³⁴ and the substance (gas or liquid) trapped inside the nano-bumps^36,37 (see Note S3, ESI†). The average h_max/R value is around 0.06, suggesting that the adhesion energy (∼4.91 mJ m⁻²) of the graphene and nano-bump system is quite small. This adhesion energy value and aspect ratio are close to that of H₂/O₂ bubbles trapped by graphene,^14,20 implying that trapped substances inside the carbon layers are gases. The dynamic evolution of nanomechanical properties of the nano-bumps was further determined by DMT modulus measurements (see Video S2, ESI†). As shown in Fig. 1g and h, the nano-bumps show darker contrast compared with the graphite (001) plane, with a DMT modulus at around 10⁸–10⁹ Pa. Although the pressure induced by the nano-bumps is neither large enough to break the flexible graphene³⁸ nor enough to liquify the trapped gases,³⁹ these “invisible” nano-bumps will inevitably block ion transport through the electrode–electrolyte interface. Overall, apart from the “visible” macroscale bubble accumulations in graphite–electrolyte interfaces, our EC-AFM measurements implied that the nano-bump formation (local delamination) is an important, but previously ignored “invisible” interfacial degradation mechanism.

1.2. Interrelated SEI formation and local delamination: nano-blister or bubble?

To understand the relation between SEI and nano-bump formation, we studied the nanoscale surface and subsurface structure evolution of an atomic step composed of a few graphene layers (Fig. S3, ESI†) upon lithiation/delithiation at different voltages using cyclic voltammetry (CV), as shown in Fig. 2a. EC-AFM was operated in the peak-force QNM mode⁴⁰ to record the surface topography and in the UFM mode to detect the subsurface mechanical structures. The technical details of UFM can be found in Note S1 (ESI†).

Fig. 2 Nanoscale SEI formation and delamination observed by operando EC-AFM and in situ nanomechanical mapping via electrochemical ultrasonic force microscopy (EC-UFM). (a) Schematic diagram of electrochemical AFM combined with ultrasonic wave excitation and the graphite atomic steps consisting of four overlapped carbon layers. (b) The cyclic voltammetry (CV) curve of the AFM EC-cell during the first lithiation in 1 M LiPF₆ in EC : DEC = 1 : 1. (c) Surface topography changes of graphite atomic steps during the first lithiation. The first top carbon layer and the second to fourth carbon layers are labelled in blue and red in Fig. 2c, No. 1. The images at different lithiation/de-lithiation states (denoted as voltage regions No. 2–9) correspond to the point in the voltage curve in Fig. 2c. (d) Surface topography and nanomechanical UFM images at open circuit potential (OCP) before CV cycles. (e) Schematic of ultrasonic wave transport through different components. Brighter UFM contrast corresponds to stiffer materials and the absence of subsurface delamination. (f) The surface topography and UFM images after the 1st cycle and (h) the 2nd cycle, and their schematic structure models (g) and (i). (j) Force indentation curve (solid line) of graphene bubble and the corresponding numerical simulation scatters by different graphene layer numbers. (k) Ultrasonic force spectra (UFS)³² (Note S2, ESI†) obtained at the bubble region, SEI component surface and graphite substrate.

Fig. 2b is the 1st CV in 1 M LiPF₆ in EC : DEC = (1 : 1 vol%) electrolyte; the voltage range can be divided into three different regions according to the different electrochemical processes that occur on the graphite surface. In region I (OCP → ∼2.25 V), the cathodic current stays constant at <1 μA (Fig. S4a, ESI†), indicating that the electrode surface is located within the thermodynamically stable voltage window. When entering the voltage region II, the typical surface-topography image recorded between 2.27 and 2.09 V can be found in Fig. 2c, in which a few scattered SEI nanoparticles (white spots) start to appear on the graphite basal plane. With the voltage decreasing to about 1.35 V, the cathodic current increases by three times. Meanwhile, more irregular nanoparticles that may be attributed to LiF start to form on the graphite basal plane² (Fig. 1c, No. 3). However, the initially monitored carbon atomic step height remains constant at around 1.01 nm (Fig. S3, ESI†), indicating that intercalation has not yet occurred in this voltage region (2.25–1.35 V). Interestingly, significant topographical changes do happen when the electrode voltage reaches voltage region III (0.80–0.65 V), in which many dense nanoparticles fully cover the electrode surface and the observed carbon step height increases to about 1.55 nm (Fig. S3, ESI†). Moreover, the first nano-bump formed in the measured area (Fig. 1c, No. 4) was also observed at this voltage region. Upon subsequent measurements (Fig. 2c, No. 5 and No. 6), SEI nanoparticles with larger size and featuring more nano-bumps continued to form on the electrode surface until the electrode current changed from cathodic to anodic. However, the surface morphology and nano-bump size stay relatively unchanged from the anodic scan to the end of the CV cycle (Fig. 2c, No. 7–9), indicating that nano-bump formation is concurrent with SEI formation and lithium intercalation.

We further introduced UFM to explore the nature of subsurface substance inside the nano-bumps. In UFM, high frequency ultrasonic excitation (∼4 MHz) with an amplitude of a few angstroms modulated at a few kHz frequency is applied from underneath the sample. The resulting cantilever deflection at modulation frequency (UFM signal) provides a measure of local mechanical properties of the sample for a wide range of elastic stiffness levels from crystalline materials to porous structures and polymers, as well as an indication of obstacles to the ultrasonic vibration propagation under the sample, such as gas or liquid bubbles. Generally, brighter UFM contrast corresponds to stiffer materials and the absence of subsurface delamination.^41,42 The UFM image is obtained by nano-tip scanning across the sample surface (Note S2, ESI†) and can be obtained simultaneously with topography as shown in Fig. 2d–i. Fig. 2d shows the topography and UFM image of carbon steps before the CV cycles, the native subsurface dislocations⁴³ and a native pore, buried underneath the top carbon layer, which appear in the UFM image as dark lines and circles, respectively. Importantly, it is also worth noting that this native pore region appears as a slightly dented surface (concave topography) as shown in 3D topography (Fig. S4(b) (ESI†)). This is significantly different from the nano-bumps generated during the electrolyte decomposition, which shows a convex topography as shown in Fig. S4(c) (ESI†). As sketched in Fig. 2e, the pore (or gas bubble) is an obstacle to ultrasonic waves, “blocking” the ultrasonic excitation coming from the substrate and therefore they appear the darkest UFM contrast compared with the SEI and graphite substrate. To study the subsurface bubbles buried under the SEI, the top SEI layer was scratched by the AFM tip before each UFM measurement (Fig. S5, ESI†). Fig. 2f and h show the UFM measurements of the carbon step after the 1st and 2nd CV cycles (see Fig. S6 (ESI†) for the DMT modulus at the same scan area). In Fig. 2f, the nano-bump (in white dashed circles) presents a significantly decreased ultrasonic response in the UFM image (α region in the red box). This indicates that the space underneath these nano-bumps has similar ultrasonic permittivity compared with the pore structure, confirming that it is a gas-filled space that has a large damping effect on the ultrasound vibration. Therefore, UFM confirms that these nano-bumps are subsurface bubbles, rather than blisters filled with liquid electrolyte. The reason is the drastic difference of acoustic impedance (product of density and sound velocity) between the solid graphite (∼3 × 10⁶ Pa s m⁻¹) and gas (∼4 × 10² Pa s m⁻¹), but a smaller difference with the impedance of liquids (∼1.5 × 10⁶ Pa s m⁻¹). It is also worth noting that the β region (at the edge of carbon steps) also shows local delamination with an increased height (Fig. 2f) but exhibits a more inhomogeneous ultrasonic response compared with the α region. This indicates that the β region may be a local delamination region filled by the liquid electrolyte and gases simultaneously.

The force–distance curves of the bubble region with vertical force modulation⁴⁴ and ultrasonic force modulation are shown in Fig. S7 and S8 (ESI†), in which one can find that the UFM response at the graphite region is about one order of magnitude larger than the UFM response at the bubble region during the indentation. Moreover, as shown in Fig. 2j, by numerical fitting of the vertical force-indentation curves during the tip indentation into the bubble described using a non-linear plate model⁴⁵ (Fig. S9 and Note S4, ESI†), we conclude that the top of this bubble consists of about 3 carbon layers, as illustrated in Fig. 2g. After the 2nd lithiation/de-lithiation cycle, the sizes of bubbles trapped inside the carbon interlayers have barely increased, while more round-shape “nano-islands” fully covering the graphite surface appear, preferentially accumulated on the step edges (Fig. 2h). These nano-islands have smaller ultrasonic responses compared with graphite but are larger than subsurface bubbles (Fig. S10, ESI†). Ultrasonic force spectroscopy (UFS) measurements performed on the bubble, nano-island and graphite substrate are shown in Fig. 2k. Under an ultrasonic excitation amplitude of 2.50 nm, the “nano-island” region exhibits a UFS response of about ∼1.08 nm, which is larger than that in the bubble areas (∼0.35 nm), but smaller than the graphite substrate (∼1.60 nm). The mechanical modulus of nano-islands determined by UFS is therefore lower than that for graphite and is around ∼10⁸–10⁹ Pa according to numerical simulation results (Fig. S11, ESI†) and close to the values of reported inorganic SEI components.^11,21,46 Therefore, as sketched in Fig. 2i, we attribute these nano-islands to stiff inorganic SEIs that seal the entrances of solvent co-intercalation, resulting in the rumination of subsurface bubble growth.

1.3. Visualizing solvent co-intercalation: the “triggers” of initial subsurface bubble formation

To understand the triggers for initial subsurface bubble formation, we focus on the interfacial nanostructure evolution of a few graphene top layers of the carbon anode under constant voltage polarization within the initial bubble formation voltage region (OCP-0.8 V vs. Li⁺/Li). The corresponding AFM results are shown in Fig. 3a. With the decrease of polarization voltage from OCP to 0.8 V, one can observe the curling of graphene layer step edges caused by the intercalation, as well as the nucleation of the “SEI seeds” on the surface. The step height of the triple-graphene layer increases from an initial value of about 0.99 nm (OCP) to 1.52 nm at 0.8 V (Fig. 3b), indicating co-intercalation of other species rather than solvated lithium ions. The value of increased height is close to the diameter of the solvent molecule, rather than the lithium–solvent complex, suggesting that the co-intercalated solvents and lithium ions stay in the de-solvation states inside the carbon layers within the measured region. According to the Raman spectrum at each polarization voltage, an increase of G peak intensity is also observed at polarization voltages 1.0 V and 0.8 V. The G peak slightly shifts to the higher wavenumber region, also indicating the increase of the force constants of the in-plane C–C bonds as a result of a small amount of solvent co-intercalation that forms the dilute stage I graphite intercalation compounds (GICs).⁴⁷

Fig. 3 The initial stage of lithium/solvent co-intercalation and SEI formation revealed by UFM. (a) Surface topography of the graphite anode under the different polarization voltages. (b) Cross-section and histogram of the step height of carbon layers during the different polarization voltages. (c) Raman spectra of the graphite surface at each polarization voltage. (d) Zoom-in ultrasonic mapping on the few-layer graphite surface at a polarization voltage of 1.0 V. (e) The 1st CV curve of a few-layer carbon step anode cycled in NaPF₆ EC : DEC = 1 : 1 electrolyte. (f)–(k) Surface topography evolution during the 1st CV cycles. The number of each image corresponds to the number labelled in Fig. 3e; the voltage in the images is the electrode voltage to Na⁺/Na.

The solvent co-intercalation was further revealed via UFM, as shown in Fig. 3d. The UFM image of the triple-layer graphene shows three different levels of contrast in the nanomechanical mapping via ultrasonic response. Compared with the UFM image before the solvent co-intercalation (Fig. S12, ESI†), we find an “electrolyte-immersed like” region near the carbon step edges (as denoted by the white dashed line in Fig. 3d). The ultrasound response inside the dashed-line area (Fig. 3d) is larger than the SEI seeds but smaller than the graphite substrate. UFS with different excitation amplitudes were recorded at the typical graphite substrate region, electrolyte-immersed region and one of the carbon step edges at the points marked as f, g and h, respectively, in Fig. 3d. The UFS results are presented in Fig. S13 (ESI†). UFS in the electrolyte-immersed region indicates lower stiffness. This region can be attributed to the solvent co-intercalated GIC, which is softer than the original graphite due to the insertion of solvent molecules between carbon layers. Moreover, UFS at the carbon step edge shows a negligible signal, confirming that the subsurface SEI/gas accumulated at the carbon edge forms a softer and lower acoustic permittivity region compared with GIC. In summary, the solvent co-intercalation and decomposition were observed to occur at the very end of the graphite edge during the initial stage of SEI/bubble formation (at ∼1 V vs. Li⁺/Li), which are the preconditions of subsurface molecular bubble formation.

Considering that only about 30 ppm water in battery electrolytes, one possible gas evolution reaction could be the two electron reduction of the co-intercalated EC solvent molecules,² which generate lithium ethylene dicarbonate (LEDC) and C₂H₂ gas at the carbon step edges. We performed a reference experiment to confirm the correlation between the co-intercalation and subsurface molecular bubble, shown in Fig. 3e–k, in which we replaced the cation Li⁺ with Na⁺ (1 M NaPF₆ in EC : DEC = 1 : 1 vol%) and evaluated the voltage-dependent electrochemical processes that occurred on a triple-layer carbon step (see Video S3 (ESI†) for the operando AFM observation) presented in Fig. 3e is the 1st CV curve during the sodiation. Compared to the CV curve in Fig. 2c for lithium electrolyte, the CV curve in Fig. 3e shows only a reduction peak corresponding to the decomposition of the electrolyte, with no oxidation peak found during the anodic scan, indicating that the reversible sodium/solvent intercalation is not present in this system. The SEI related reduction peak almost disappeared during the second cycle (Fig. S14, ESI†), suggesting that the electrolyte decomposition mainly occur on the anode surface during the first CV cycle. The surface topography images in Fig. 3 panels (f)–(k) revealed that the onset of SEI formation occurs at around 1.1 V, while the carbon atomic step height does not change, step edge curls do not appear, whereas SEI accumulation at the carbon atomic step was observed (Fig. S15 and S16, ESI†). To summarise, if there is no cation/solvent co-intercalation, even when the electrode surface is polarized beyond the electrochemical stability window of the electrolyte, the subsurface “molecular bubbles” cannot be generated.

1.4. Co-intercalation and decomposition of Li⁺(EC)_x(DEC)_y—effects of intermolecular interactions

Graphite–electrolyte interfacial degradation, either resulting in surface SEI or subsurface molecular bubble formation, is dominated by the interfacial chemistry of lithium–solvation complexes. In EC/DEC electrolytes, the coordination number of lithium ions is around 4–5, and Li⁺(EC)₄, Li⁺(EC)₃(DEC)₁ and Li⁺(EC)₂(DEC)₂ are the three most probable complexes, but which of them dominates the interfacial degradation remains elusive. Fourier transform infrared (FTIR) spectroscopy and density functional theory (DFT) modelling were used to find the specific solvation complexes that trigger the subsurface molecular bubbles.

First of all, we tuned the ratio of Li⁺(EC)₄, Li⁺(EC)₃(DEC)₁ and Li⁺(EC)₂(DEC)₂ by varying the EC : DEC ratios in 1 M LiPF₆ in EC/DEC electrolytes; FTIR spectroscopy was then performed as shown in Fig. 4a and b (full spectra can be found in Fig. S17, ESI†) to confirm the ratio of these solvation complexes. In the C–O band region of DEC (Fig. 4a), we observed a non-negligible proportion of a solvated C–O band (∼1302 cm⁻¹) of DEC, confirming that the DEC molecules participate in the solvation with Li⁺.⁴⁸ With the increase of the EC ratio, the peak intensity of solvated DEC decreased from about 37.8% to 23.8%, indicating a decrease of Li⁺(EC)₃(DEC)₁ and Li⁺(EC)₂(DEC)₂ in the electrolyte. By contrast, the C–O peak intensity of the solvated EC was slightly enhanced (Fig. 4b), indicating a much stronger and more stable Li⁺–EC coordination interaction compared with Li⁺–DEC. With the increasing EC ratio from EC : DEC = 4 : 6 to EC : DEC = 7 : 3, the FTIR peak intensity of solvated-EC vs. solvated-DEC increased by about 64%, confirming the abundance of the Li⁺(EC)₄ solvation complex in the electrolyte, with a high EC ratio. In this high-Li⁺(EC)₄ content electrolyte, we observed more profound graphite surface delamination by operando AFM (Fig. S18, ESI†), supporting the proposition that the higher proportion of Li⁺(EC)₄ in the electrolyte may cause more solvent co-intercalation-induced interfacial degradation.

Fig. 4 Effects of dipole–dipole interaction on the co-intercalation and reductive decomposition of lithium–solvent complexes. The C–O band of (a) DEC and (b) EC FTIR spectra in 1 M LiPF₆ electrolyte with various EC : DEC ratios. (c) The side-view and top view of the structure model of three types of solvation complexes, Li⁺(EC)₄, Li⁺(EC)₃(DEC)₁ and Li⁺(EC)₂(DEC)₂, adsorbing on the graphite edge planes. (d) Adsorption energy and de-solvation energy of three types of solvation complexes on the graphite edge plane. (e) 3D representation of the HOMO/LUMO energy level of Li⁺(EC)₄ interacted with a DX and FB free-solvent molecule.

Importantly, we also noticed that there is an evident blueshift of the C–O peak of solvated DEC (∼1303 cm⁻¹), which can be attributed to the enhanced dipole–dipole interaction of solvated DEC with free EC molecules.⁴⁹ These dipole–dipole interactions were also observed in the EC/DEC-mixed solvent without the lithium salt (Fig. S19, ESI†). It has been reported that dipole–dipole interactions between the free-solvents and solvated solvents play a significant role in battery interfacial chemistry.^49,50 Given that the polarizing effect of Li⁺ on the EC/DEC molecules can further enhance these dipole–dipole interactions between free-solvent and Li⁺(EC)_x(DEC)_y and affect their electrochemical behaviours,⁵¹ we considered these interactions to evaluate the adsorption energy, chemical stability and de-solvation energy of Li⁺(EC)_x(DEC)_y using DFT calculations. As shown in Fig. 4c, we constructed the models with the Li-solvation complexes, Li⁺(EC)₄, Li⁺(EC)₃(DEC)₁ and Li⁺(EC)₂(DEC)₂, absorbing on the graphite edge plane to evaluate their adsorption and de-solvation energies. The intermolecular interactions between the free solvents and lithium–solvation complexes were treated by an implicit model by tuning the dielectric constant of the simulation system. The dielectric constants were swept from 10 to 100 according to the changes of EC : DEC ratios⁵² in the electrolytes. As shown in Fig. 4d, in the solvent environment with various EC : DEC ratios, Li⁺(EC)₄ always shows the lowest energy of adsorption onto the graphite surface, but the highest de-solvation energy. This indicates that for Li⁺(EC)₄ it is energetically favourable to adsorb on the graphite edge planes and then co-intercalate into the graphite lattice without the de-solvation process. By contrast, Li⁺(EC)₃(DEC)₁ and Li⁺(EC)₂(DEC)₂ exhibit relatively high adsorption energies on the graphite surface, indicating that they are statistically distributed further away from the graphite electrode surface. However, once they approach the graphite surface, they tend to desolvate before entering the graphite lattice without solvent co-intercalation. This further suggests that Li⁺(EC)₄ is the culprit for the initial solvent co-intercalation and graphite delamination, confirming the AFM observations in Fig. S18 (ESI†).

According to the DFT calculations in Fig. 4d, it can be noted that modifying the dielectric environment will allow control over the de-solvation processes of Li⁺(EC)_x(DEC)_y. It has also been reported that the de-solvation energy of lithium complexes can be tailored using weakly/non-solvating diluents, which has been employed as a promising strategy to inhibit solvent co-intercalation into graphite, as well as to enhance graphite–electrolyte interfacial stability.^53,54 Hence, as a proof of concept, weakly-solvating (1,4 dioxane, DX) and non-solvating solvents (fluorobenzene, FB) with dielectric constants of about 2.25 and 5.42, respectively, were chosen as the diluents to test the ability to suppress the Li⁺(EC)₄ co-intercalation. Fig. 4e shows focused analysis of the lowest unoccupied molecular orbital (LUMO) and highest occupied molecular orbital (HOMO) of Li⁺(EC₄) with additional consideration of the intermolecular interactions from the free solvents. A DX and a FB were attached to Li⁺(EC)₄ to represent the intermolecular interaction of free solvents in the electrolytes with DX and FB diluents. As shown in Fig. 4e, the LUMO orbital remains in one of the solvated EC molecular orbitals, and the additional DX or FB molecules can elevate the LUMO energy level of Li⁺(EC)₄⁵⁵ by about 0.28 eV and 0.14 eV, respectively. Moreover, the higher the LUMO orbital energy of the solvated EC is, the weaker the intermolecular interactions between the solvated EC and the lithium ion.⁵⁶ This indicates that DX and FB have both weakened the intermolecular interaction between EC and Li⁺ inside Li⁺(EC)₄, which can reduce its de-solvation energy and facilitate the pure lithium intercalation into the graphite anode. Additionally, this aside, the HOMO orbital fully transfers to the additive DX/FB molecules in Li⁺(EC)₄ + DX (−8.61 eV) and Li⁺(EC)₄ + FB (−9.14 eV), and the HOMO levels are still sufficiently low to resist oxidation reactions on the cathode surface.

1.5. Inhibiting the subsurface gas evolution via interfacial molecular structure optimization

To verify the effect of DX and FB diluents experimentally, we conducted the galvanostatic charge/discharge measurements of a graphite//Li half-cell using 1 M LiPF₆ in EC and DEC containing FB/DX diluents (Fig. 5a). We found that fully replacing the DEC by weakly solvating DX results in serious solvent co-intercalation and graphite delamination, characterised by a long discharge voltage platform at around 0.9 V. This is due to the existence of a large amount of the strongly solvated Li⁺(EC)₄ in the electrolyte, emphasizing the presence of competing solvating effects of DEC as the co-solvent. When only about 23 mol% of the solvent was replaced by DX (EC : DEC : DX = 5 : 5 : 3), solvent co-intercalation was effectively alleviated, but unwanted electrolyte decomposition remains, showing a lower ICE (∼57%) compared with the EC : DEC = 1 : 1 electrolyte (∼72%). The low ICE is due to the effect of non-negligible Li-DX coordination, facilitating the DX solvent to enter the primary solvation shell (Fig. S20, ESI†) and modifying the reductive capability of Li⁺(EC/DEC/DX)_x solvation complexes. Significantly, we found that adding FB solvent can enhance the ICE effectively (Fig. S21, ESI†), and when about 23 mol% of FB was used as the diluent, not only was solvent co-intercalation suppressed, but also electrolyte decomposition, showing a remarkably higher ICE of ∼93% (Fig. 5a). The high ICE probably derives from the moderate dipole–dipole interaction between the C–F bonds of FB and the C–H of EC molecules⁶¹ (see FITR in Fig. S22, ESI†), weakening the coordination effect between EC and Li⁺, therefore allowing the de-solvation energy of Li⁺(EC)₄ to be reduced. As a result, the electrode surface stays relatively smooth after lithiation in the EC : DEC : FB = 5 : 5 : 3 electrolyte (Fig. S18d, ESI†), in which subsurface bubbles (Fig. S18b, ESI†) and local delamination (Fig. S18c, ESI†) were barely observed. It is worth mentioning that further increasing the ratio of FB solvent could result in the precipitation of salt; this is due to the low-dielectric permittivity of FB solvent that shows limited solubility toward the LiPF₆ salt. Fig. 5b and c show the simulated radial distribution function and coordination number of lithium ions with the solvents and PF₆⁻ anion. These figures reveal that in the optimized electrolyte, no FB molecules reside within the first solvation shell (<0.3 nm), while the ratio of PF₆⁻ inside the first solvation shell is twice that of the electrolyte without FB diluent. Consequently, the number of contact ion pairs increases in the electrolyte containing FB diluent compared to the electrolyte without it. This indicates that FB diluent can enhance Li–anion interaction and reduce Li–EC interaction by optimizing the dipole–dipole interaction between solvents.

Fig. 5 Electrochemical performance and interfacial structure of graphite–electrolyte with and without diluent optimization. (a) Formation charge–discharge curves of the graphite anode in 1 M LiPF₆ in EC : DEC = 1 : 1, EC : DX = 1 : 1, EC : DEC : DX = 5 : 5 : 3 and EC : DEC : FB = 5 : 5 : 3 electrolyte (molar ratio). The radial distribution function and coordination numbers of lithium ions with the solvents and anion in the electrolyte without (b) and with (c) FB diluent. Force–distance-curves (FDC) measured on a graphite anode surface in (d) EC : DEC = 1 : 1 and (e) EC : DEC : FB = 5 : 5 : 3 electrolyte. (f) DFT calculation model for the adsorption energies of EC, DEC and FB molecules on the graphite surface. X-ray photoelectron spectroscopy (XPS) milling spectra of the F 1s peak on the graphite electrode surface after formation using (g) EC : DEC = 1 : 1 electrolyte without FB diluent and (h) optimized electrolyte with FB (EC : DEC : FB = 5 : 5 : 3) Schematic diagram of structural degradation and interfacial chemistry graphite anode cycled in the electrolyte without (i) and with (j) the FB additive. (k) Formation charge/discharge curves of a lithium cobalt oxide (LCO)/graphite full-cell using EC : DEC = 1 : 1 electrolyte without FB diluent and optimized electrolyte with FB (EC : DEC : FB = 5 : 5 : 3) at 0.05C and 0.5C.

Apart from the “passivation” and “dragging” effects of FB molecules toward the Li⁺(EC)₄ complex, we also observed the preferential adsorption of FB molecules on the graphite surface. As shown in Fig. 5d and e, interfacial molecular structures near the graphite surface in the electrolytes with and without FB diluent were measured by AFM-based force–distance spectroscopy⁵⁷ (Note S5 and Fig. S23, ESI†). In the conventional electrolyte without FB diluent (Fig. 5d), we observed the adhesive force region (at around 0.31–1.88 nm) as a result of the competitive effects of solvation force and Derjaguin–Landau–Verwey–Overbeek (DLVO) force,^58,59 followed by a repulsive force region at the inner electrical double layer (EDL) due to the confined effect of the tip and sample surface.^57,60 However, the adhesive force region disappeared in the electrolyte with FB diluent (Fig. 5e). After adding FB, the disappearance of the adhesive force region is due to the change in the specific adsorption of ions/molecules on the graphite surface, which modified interfacial dielectric permittivity. This unambiguously points to the localized preferential adsorption and accumulation of low-dielectric-constant FB molecules on the graphite surface; therefore, the anode–electrolyte interfacial compatibility can be greatly affected by this disproportionate distribution of solvent near the electrode surface.⁶¹ The preferential accumulation of FB on the graphite–electrolyte interface was also confirmed by DFT calculations (Fig. 5f), in which we found that FB has a lower adsorption energy of −0.48 eV on the graphite surface compared to EC (−0.4 eV). The preferential accumulation of a low-dielectric FB and DEC molecular layer may act as a protective molecular layer preventing solvent co-intercalation and decomposition.

Additionally, the preferential adsorption of FB on the graphite–electrolyte surface may result in the formation of an SEI layer with stable metallic-fluorine species.^62,63 This is because that the FB diluent enhanced Li⁺–PF₆⁻ interaction, which facilitates the PF₆⁻ to enter the first solvation shell (Fig. 5c) and help to form an anion-derived SEI layer.^64–66 The anion-derived SEI with an inorganic–fluoride rich species was confirmed by the X-ray photoelectron spectroscopy (XPS) milling spectra in Fig. 5g and h. As shown in the figures, with the increase of milling depth, the F 1s peak intensity at around 685.5 eV (attributed to LiF⁶⁷) decreases on the graphite surface cycled in FB-free electrolyte (Fig. 5g) but increases on the graphite surface cycled in electrolyte with FB additive (Fig. 5h).

According to the comprehensive characterization, we summarized the key understanding of solvent co-intercalation, subsurface gas evolution and SEI formation behaviours on the graphite–electrolyte interface in the electrolyte with and without the FB additive as shown in Fig. 5i and j. As sketched in the figures, the FB molecules preferentially adsorb on the graphite surface, participating in reducing the de-solvation energy of Li⁺, which effectively inhibits the Li⁺(EC)₄ co-intercalation and decomposition within the graphite interlayers. As a result, the subsurface molecular bubbles or other types of micro-structural damage disappeared in the graphite anode matching with the optimized FB electrolyte. This optimized FB-contained electrolyte has special interfacial molecular structures that can also facilitate the formation of a stable inorganic-rich SEI layer. Benefiting from the reduced lithium de-solvation energy and optimized interfacial chemistry, the formation charge/discharge curves of the graphite/LCO cells using EC/DEC-based electrolyte with FB diluent greatly exceeds that of cells without FB diluent as shown in Fig. 5k. At 0.05C formation, the solvent co-intercalation voltage occurring at around 3.5–3.8 V disappeared after adding FB diluent. Moreover, the fluctuations in the formation charge/discharge curves are also smoothed after adding FB diluent at 0.5C formation, suggesting that time and energy-consuming formation processes can be avoided in the optimized electrolyte. More importantly, we fabricated the graphite/LCO pouch cell (4 Ah) to evaluate the capability of FB additives to accelerate the formation process of high mass loading electrodes (Fig. S24, ESI†). Although the coulombic efficiencies of these pouch cells are not excellent due to the high electrode mass loading (about 19.4 mg cm⁻² for the cathode and 10.1 mg cm⁻² for the anode (single-side)), it is evident that the pouch cell with the FB additive after formation at a large current density (0.5C) shows even better cycle capacity retention compared with the cell using a small formation rate (0.2C) in the electrolyte without the FB additive. This holds significant promise for reducing battery formation time while preserving a robust and high-quality SEI surface passivation layer, thereby lowering the overall costs of secondary lithium/sodium-ion battery fabrication.

2. Conclusion

In this work, we introduced ultrasonic-based EC-AFM to study the nanoscale SEI formation and subsurface gas evolution of a graphite anode during lithiation/de-lithiation in EC/DEC-based electrolyte. We observed subsurface gas evolution, starting simultaneously with solvent co-intercalation (<∼0.1 V vs. Li⁺/Li) during lithiation, resulting in local delamination and the formation of “subsurface molecular bubbles” trapped between the carbon layers. The nanostructures and nano-mechanical properties of the formed SEI particles and molecular bubbles were systematically explored by ultrasonic force spectroscopy, which revealed that the solvent co-intercalation and decomposition at edge planes may be the two preconditions for the formation of subsurface molecule bubbles.

Focusing on solvent co-intercalation and decomposition, we combined implicit and explicit models to study the weak interactions between the free solvent and the Li⁺(EC)₄ solvation complex using DFT calculations. Our DFT calculations, by taking into account the intermolecular interactions of free solvent outside the primary solvation shell in EC/DEC electrolytes, suggest that the specific complex involved in co-intercalation is Li⁺(EC)₄, possessing high de-solvation energy and low adsorption energy on the graphite edge plane. Leveraging the understanding of interfacial degradation in conventional EC/DEC electrolytes, FB diluent was introduced to reduce the cation–solvent (ion–dipole) interaction of Li⁺(EC)₄, enabling pure cation intercalation and enhancing chemical stability. The graphite electrode surface lithiation in the optimized electrolyte is smooth and clean, without interfacial structural degradation, facilitating enhanced cycle stability and formation current density of LCO/Graphite full cells.

Overall, our ultrasonic-based EC-AFM nanoscale characterization studies aim to uncover the underlying/root causes for the formation of gases within the carbon layers – ‘the hidden subsurface bubbles’ clearly indicate that the formation of nano-bumps/delamination and the subsequent inaccessible lithium storage site contribute to the initial capacity loss in addition to SEI layer formation. This phenomenon not only enhances our understanding of the interfacial degradation mechanism in Li ion batteries, but also holds promise for translation to other batteries. Although substantial research has been carried out to unravel the formation, storage and cycling degradation mechanism through invasive and non-invasive techniques, these insights through nanoscale characterization techniques underpin the importance of yet another perspective that could enrich our understanding of the battery degradation mechanism.

3. Methods

3.1. Sample preparation

We used a commercial graphite anode (BTR company, China) with an average particle size of about 15–25 μm for the subsurface bubble observation on the basal planes of graphite particles during galvanostatic lithiation. The graphite anode was mixed with 10 wt% polyvinylidene fluoride (PVDF), 10 wt% super P carbon conductive and N-methyl-2-pyrrolidone (NMP) to prepare the slurry. The obtained slurry was doctor-blade coated onto the copper foil surface and dried overnight before cutting to the target shape for the EC-AFM or Raman measurements. The mass loading of single-side coated electrodes is about 20 mg cm⁻² for the cathode and 10 mg cm⁻² for the anode electrode, respectively, which is controlled to make sure the N : P ratio and the mass loading is close to those of the double-side coated electrodes used in pouch-cellss. This graphite anode was also used to evaluate the electrochemical performance of different electrolytes in a half-cell using a CR2032 coin-cell formfactor, as well as a full cell in a three-electrode electrochemical cell. A LiCoO₂ cathode (Shanshan company, China) and lithium metal were used as working electrodes and reference electrodes, respectively. Few-layer graphene atomic steps (with 3–4 carbon layers) obtained from the HOPG (438HP-AB, SPI-1 Grade) were selected as the model sample to study solvent co-intercalation and SEI formation. Before EC-AFM measurements, the graphene atomic step surface was cleaned by nano-scratching in AFM contact mode with a tip force of about 20 nN. The LiPF₆ salt was dissolved in the solvent mixed with different EC : DEC : FB/DX ratios overnight inside a glovebox to obtain a transparent electrolyte.

The pouch cells were fabricated in the commercial production line in EBTEB Electronics Co. Ltd. In the LCO//graphite pouch cell, the cathode typically consists of about 97% LiCoO₂, 1.7% carbon black, and 1.3% PVDF binder, and the anode is composed of about 96% graphite, 1.3% conductive additive, and 2.7% carboxymethyl cellulose (CMC) binder. The solvents used for cathode and anode electrodes are NMP and DI water, respectively. The current collectors used for the cathode and anode are 10 μm Al foil and 10 μm Cu foil, respectively. We applied calendaring pressures of about 80 MPa for both cathode and anode electrodes to obtain the total thickness of about 105 ± 3 μm and 124 ± 3 μm for the double-side coated region, respectively. This yields a single-side mass loading of about 19.4 mg cm⁻² for the cathode and 10.1 mg cm⁻² for the anode. For a 4 Ah pouch cell, the typical amount of electrolyte used is about 2.0 g Ah⁻¹; the fabricated pouch cell unit (including the control circuits and outside plastic package) has an average weight of about 102 g, corresponding to about 145 Wh kg⁻¹ (measured at 1C and using a middle value voltage of 3.7 V vs. Li⁺/Li). The typical formation protocol for the pouch cell (without the FB additive) involves an initial charge at a low C-rate (0.2C) to a cut-off voltage of 3.2 V, 3.6 V, 4.1 V and 4.2 V, followed by a resting period (120 s) to allow the formation of a SEI on the anode. The cell is then discharged to 3.0 V at 0.2C. This process is repeated for two charge/discharge cycles to stabilize the cell, ensuring proper activation of the electrodes and long-term performance. The protocol is typically carried out at room temperature (around 25 °C).

3.2. Electrochemical AFM, UFM, Raman and XPS characterization

The anode/cathode were mounted using Torr Seal epoxy on a Dimension Icon electrochemical cell (Bruker, USA). Cell open circuit potential was recorded for 300 seconds after the injection of electrolyte (∼0.5 mL), and then about 200 nN force was applied to the tip to clean the electrode surface by scanning in a 15 × 15 μm² area before the operando EC-AFM measurements. A Reference 600+ potentiostat (Gamry, USA) was used for cyclic voltammetry (CV) and galvanostatic charge/discharge measurements. The C-rate of the graphite anode is calculated based on the specific capacity of 372 mAh g⁻¹. CV data were obtained at a scanning rate of 0.5–0.7 mVs⁻¹ between OCP and 0.01 V vs. Li⁺/Li in lithium-ion battery electrolytes (or Na⁺/Na in sodium ion battery electrolytes). 1 M LiPF₆ in ethylene carbonate and dimethyl carbonate (EC : DEC) = 1 : 1 (v : v) was used as the regular electrolyte, to compare with the other electrolytes with various solvent ratios. Operando electrochemical SPM was performed inside the glovebox (MBRAUN, Germany) with oxygen and moisture content <0.5 ppm. Nano-mechanical properties were measured by ultrasonic force microscopy (UFM) (Note S1, ESI†) and peak force quantitative nanomechanical mapping (QNM) in a liquid environment. The tips used were Bruker ScanAsyst-fluid⁺, force modulation and non-contact (Apex Probes) with force constants of 0.5 N m⁻¹, 3 N m⁻¹ and 40 N m⁻¹, respectively. The amplitude modulation force spectra of electrical double-layer structures were detected using a hard magnetic tip (NSC36, Micromash). The calibration methods of the effective viscosity using tapping amplitude and phase vs. distance curves can be found in Note S5 (ESI†). The bulk-state solvation structures and intermolecular interactions of different electrolytes were measured using a FTIR spectrometer (Thermo Scientific Nicolet iS50). An operando Raman spectrometer (LabRAM HR Evolution) using a green laser with a 532 nm wavelength was employed to measure the ion-intercalation into graphite anodes and few-layer graphene assembled in an EL-CELL (ECC-Opto-Std, Germany). XPS measurements were conducted using a Kratos Analytical AXIS Supra spectrometer equipped with a monochromatic Al K_α (1486.7 eV) X-ray source, operating at 15 kV and 15 mA, and an electron gun for charge neutralization. Depth profiling was performed in situ using an Ar Gas Cluster Ion Source (GCIS; Kratos Analytical Inc., Minibeam 6). The ion beam utilized a raster size of 3  ×  3 mm². All spectra were analyzed using CASA XPS software.

3.3. DFT and MD simulations

DFT calculations were applied using the Vienna Ab Initio Simulation Package (VASP)⁶⁸ and Gaussian16 package.⁶⁹ VASPsol,^70,71 a package that incorporates solvation into VASP within a self-consistent continuum model, was used for implicit solvation calculations with growing dielectric constant referring to higher EC density. We used VASP and VASPsol in the adsorption energy and de-solvation energy calculations, where the PBE functional⁷² was used and the energy cut-off was set at 500 eV with a gamma-only k-point mesh. The graphite edge was modelled by 6 layers of a 3 × 8 graphite sheet with an armchair edge. The adsorption energy E_ads was given by

E_ads = E_tot −(E_comp + E_graphite)

where E_tot refers to the DFT calculated total energy of adsorbed complexes/molecules on graphite, E_comp refers to the energy of complexes/molecules in vacuum and E_graphite refers to the energy of the graphite edge support. The desolvation energy E_desol was given by

E_desol = (E_mol + E_Li⁺) − E_comp

where E_mol refers to the energy of electrolyte molecules in a vacuum and E_Li⁺ is the energy of the Li-ion. We used Gaussian16 for molecular orbital calculation, where the B3LYP exchange function⁷³ along with a 6-311G basis set was used to balance the accuracy and the computational resource. All the structures were fully relaxed with a DFT-D3⁷⁴ (with BJ damping) correction before performing energy and electronic structure calculations.

MD simulations of electrolytes with/without FB were performed using GROMASC software.⁷⁵ The temperature was controlled at 300 K using the Nose–Hoover thermostat. The many-body force field OPLS-AA was used.⁷⁶ The partial atomic charges of PF₆⁻, FB, and EC/DEC are obtained from LigParGen⁷⁷ and Automated Force Field Topology Builder (ATB).⁷⁸ The electrostatic interactions were computed using the Particle Mesh Ewald (PME) method. The cut-off distance of 1.5 nm was adopted for electrostatics and van der Waals interactions. The polarization of the electrode is employed as demonstrated in our previous work.¹¹ The simulation boxes were 5.6 × 5.5 × 11.0 nm³ with 3 layers of the graphene sheet as the carbon anode. In the simulation boxes, the reference electrolyte consists of 156 LiPF₆, 1000 EC and DEC molecules. The FB added electrolyte consists of 156 LiPF₆, 770 EC and DEC, 460 FB molecules corresponding to the molar concentration and solvent ratio in the two electrolytes under ambient conditions. The equilibrium states of the system were reached by doing an energy optimization and then an NVT simulation of 15 ns with 1 fs for each step. The RDF and coordination number analyses were employed for the trajectory of last 1 ns.

Author contributions

Y. C. and O. K. developed the overall EC-AFM and UFM methodology and carried out the data analysis. Y. C., W. Z. and M. N. performed the optical microscopy, Raman, and FTIR measurements. Y. N., G. Z., S. M., D. W., and C. G. provided the commercial electrodes and customized electrolytes. Y. C., J. T., J. L., L. Z. and O. K. designed and performed the electrochemical measurements. W. X. and H. C. performed the computational studies. All authors contributed to the manuscript preparation and revision. Y. L., H. C., O. K., and Z. H. supervised the work.

Conflicts of interest

The authors declare no competing interests.

Data availability

Data for this article, including AFM images, force curves and MD models, are available at Figshare at: https://doi.org/10.6084/m9.figshare.28366691.

Acknowledgements

The authors wish to acknowledge the financial support from the Natural Science Foundation of China (No. 62474041, 52403294, 52172243 and 22179020), the Science Foundation of the Fujian Province (No. 2023J01521), Fujian Province Technology Innovation Key Research and Industrialization project (2024XQ007), Foreign Science Technology Cooperation Project of Fuzhou Science and Technology Bureau (2024-Y-006), the Faraday Institution (NEXGENNA grant number FIRG018), EU Graphene Flagship Core 3 project and EPSRC project EP/V00767X/1. We are also grateful to Bruker UK, Leica Instruments, and LMA Ltd. The authors also acknowledge scientific insights by the NEXGENNA consortium in the new methodology development. The authors also acknowledge scientific insights by Matthew Dyer. W. X. and H. C. acknowledge National Science and Technology Council, NSTC (111-2221-E-007-087-MY3, 111-2112-M-007-028-MY3) in Taiwan for their financial support. The computational resources were supported by TAIWANIA at the National Center for High-Performance Computing (NCHC) of National Applied Research Laboratories (NARLabs) in Taiwan.

References

1. D. L. Wood, J. Li and S. J. An, Joule, 2019, 3, 2884–2888.
2. T. Liu, L. Lin, X. Bi, L. Tian, K. Yang, J. Liu, M. Li, Z. Chen, J. Lu, K. Amine, K. Xu and F. Pan, Nat. Nanotechnol., 2019, 14, 50–56.
3. Y. Xiang, M. Tao, X. Chen, P. Shan, D. Zhao, J. Wu, M. Lin, X. Liu, H. He, W. Zhao, Y. Hu, J. Chen, Y. Wang and Y. Yang, Nat. Commun., 2023, 14, 177.
4. P. Liu, L. Yang, B. Xiao, H. Wang, L. Li, S. Ye, Y. Li, X. Ren, X. Ouyang, J. Hu, F. Pan, Q. Zhang and J. Liu, Adv. Funct. Mater., 2022, 32, 2208586.
5. S. J. An, J. Li, C. Daniel, D. Mohanty, S. Nagpure and D. L. Wood, Carbon, 2016, 105, 52–76.
6. S. Tu, B. Zhang, Y. Zhang, Z. Chen, X. Wang, R. Zhan, Y. Ou, W. Wang, X. Liu, X. Duan, L. Wang and Y. Sun, Nat. Energy, 2023, 8, 1365–1374.
7. S. S. Zhang, Front. Energy Res., 2014, 2, 59.
8. B. Michalak, B. B. Berkes, H. Sommer, T. Bergfeldt, T. Brezesinski and J. Janek, Anal. Chem., 2016, 88, 2877–2883.
9. X. Liu, L. Yin, D. Ren, L. Wang, Y. Ren, W. Xu, S. Lapidus, H. Wang, X. He, Z. Chen, G.-L. Xu, M. Ouyang and K. Amine, Nat. Commun., 2021, 12, 4235.
10. P. G. Kitz, M. J. Lacey, P. Novák and E. J. Berg, J. Power Sources, 2020, 477, 228567.
11. Y. Chen, W. Wu, S. Gonzalez-Munoz, L. Forcieri, C. Wells, S. P. Jarvis, F. Wu, R. Young, A. Dey, M. Isaacs, M. Nagarathinam, R. G. Palgrave, N. Tapia-Ruiz and O. V. Kolosov, Nat. Commun., 2023, 14, 1321.
12. A. Schiele, B. Breitung, T. Hatsukade, B. B. Berkes, P. Hartmann, J. Janek and T. Brezesinski, ACS Energy Lett., 2017, 2, 2228–2233.
13. H. Teshima, Q.-Y. Li, Y. Takata and K. Takahashi, Phys. Chem. Chem. Phys., 2020, 22, 13629–13636.
14. H. C. Ko, W. H. Hsu, C. W. Yang, C. K. Fang, Y. H. Lu and I. S. Hwang, Langmuir, 2016, 32, 11164–11171.
15. L. Zhang, C. Tsolakidou, S. Mariyappan, J.-M. Tarascon and S. Trabesinger, Energy Storage Mater., 2021, 42, 12–21.
16. W. Du, R. E. Owen, A. Jnawali, T. P. Neville, F. Iacoviello, Z. Zhang, S. Liatard, D. J. L. Brett and P. R. Shearing, J. Power Sources, 2022, 520, 230818.
17. D. Goers, M. Holzapfel, W. Scheifele, E. Lehmann, P. Vontobel and P. Novák, J. Power Sources, 2004, 130, 221–226.
18. Z. Deng, Z. Huang, Y. Shen, Y. Huang, H. Ding, A. Luscombe, M. Johnson, J. E. Harlow, R. Gauthier and J. R. Dahn, Joule, 2020, 4, 2017–2029.
19. A. J. Louli, A. Eldesoky, R. Weber, M. Genovese, M. Coon, J. deGooyer, Z. Deng, R. T. White, J. Lee, T. Rodgers, R. Petibon, S. Hy, S. J. H. Cheng and J. R. Dahn, Nat. Energy, 2020, 5, 693–702.
20. H. An, B. H. Tan, J. G. S. Moo, S. Liu, M. Pumera and C. D. Ohl, Nano Lett., 2017, 17, 2833–2838.
21. Z. Zhang, K. Smith, R. Jervis, P. R. Shearing, T. S. Miller and D. J. L. Brett, ACS Appl. Mater. Interfaces, 2020, 12, 35132–35141.
22. H.-Y. Song and S.-K. Jeong, J. Power Sources, 2018, 373, 110–118.
23. R. A. House, U. Maitra, M. A. Perez-Osorio, J. G. Lozano, L. Jin, J. W. Somerville, L. C. Duda, A. Nag, A. Walters, K. J. Zhou, M. R. Roberts and P. G. Bruce, Nature, 2020, 577, 502–508.
24. R. A. House, H. Y. Playford, R. I. Smith, J. Holter, I. Griffiths, K.-J. Zhou and P. G. Bruce, Energy Environ. Sci., 2022, 15, 376–383.
25. D. Shin, J. B. Park, Y. J. Kim, S. J. Kim, J. H. Kang, B. Lee, S. P. Cho, B. H. Hong and K. S. Novoselov, Nat. Commun., 2015, 6, 6068.
26. X. Gao, B. Li, K. Kummer, A. Geondzhian, D. A. Aksyonov, R. Dedryvère, D. Foix, G. Rousse, M. Ben Yahia, M.-L. Doublet, A. M. Abakumov and J.-M. Tarascon, Nat. Mater., 2025, 24, 743–752.
27. H. Huo, K. Huang, W. Luo, J. Meng, L. Zhou, Z. Deng, J. Wen, Y. Dai, Z. Huang, Y. Shen, X. Guo, X. Ji and Y. Huang, ACS Energy Lett., 2022, 7, 650–658.
28. W. Xu, Y. Yang, F. Shi, L. Li, F. Wen and Q. Chen, Cell Rep. Phys. Sci., 2023, 4, 101579.
29. S. Stock, F. Diller, J. Böhm, L. Hille, J. Hagemeister, A. Sommer and R. Daub, J. Electrochem. Soc., 2023, 170, 060539.
30. Y. Chen, S. Zhang, W. Zhang, A. Quadrelli, S. Jarvis, J. Chen, H. Lu, N. Mangayarkarasi, Y. Niu, J. Tao, L. Zhang, J. Li, Y. Lin, Z. Huang and O. Kolosov, Appl. Phys. Rev., 2024, 11, 021422.
31. F. Dinelli, S. K. Biswas, G. A. D. Briggs and O. V. Kolosov, Phys. Rev. B: Condens. Matter Mater. Phys., 2000, 61, 13995–14006.
32. B. J. Robinson and O. V. Kolosov, Nanoscale, 2014, 6, 10806–10816.
33. S. Weng, G. Yang, S. Zhang, X. Liu, X. Zhang, Z. Liu, M. Cao, M. N. Ates, Y. Li, L. Chen, Z. Wang and X. Wang, Nanomicro Lett., 2023, 15, 215.
34. E. Khestanova, F. Guinea, L. Fumagalli, A. K. Geim and I. V. Grigorieva, Nat. Commun., 2016, 7, 12587.
35. D. A. Sanchez, Z. Dai, P. Wang, A. Cantu-Chavez, C. J. Brennan, R. Huang and N. Lu, Proc. Natl. Acad. Sci. U. S. A., 2018, 115, 7884–7889.
36. P. Jia, W. Chen, J. Qiao, M. Zhang, X. Zheng, Z. Xue, R. Liang, C. Tian, L. He, Z. Di and X. Wang, Nat. Commun., 2019, 10, 3127.
37. H. Ghorbanfekr-Kalashami, K. S. Vasu, R. R. Nair, F. M. Peeters and M. Neek-Amal, Nat. Commun., 2017, 8, 15844.
38. T. Cui, S. Mukherjee, P. M. Sudeep, G. Colas, F. Najafi, J. Tam, P. M. Ajayan, C. V. Singh, Y. Sun and T. Filleter, Nat. Mater., 2020, 19, 405–411.
39. G. Zamborlini, M. Imam, L. L. Patera, T. O. Mentes, N. Stojic, C. Africh, A. Sala, N. Binggeli, G. Comelli and A. Locatelli, Nano Lett., 2015, 15, 6162–6169.
40. R. Kumar, A. Tokranov, B. W. Sheldon, X. Xiao, Z. Huang, C. Li and T. Mueller, ACS Energy Lett., 2016, 1, 689–697.
41. A. Ben Gouider Trabelsi, F. V. Kusmartsev, B. J. Robinson, A. Ouerghi, O. E. Kusmartseva, O. V. Kolosov, R. Mazzocco, M. B. Gaifullin and M. Oueslati, Nanotechnology, 2014, 25, 165704.
42. R. Mazzocco, B. J. Robinson, C. Rabot, A. Delamoreanu, A. Zenasni, J. W. Dickinson, C. Boxall and O. V. Kolosov, Thin Solid Films, 2015, 585, 31–39.
43. K. Yamanaka, H. Ogiso and O. Kolosov, Appl. Phys. Lett., 1994, 64, 178–180.
44. R. Garcia, Chem. Soc. Rev., 2020, 49, 5850–5884.
45. A. Castellanos-Gomez, M. Poot, G. A. Steele, H. S. J. van der Zant, N. Agrait and G. Rubio-Bollinger, Adv. Mater., 2012, 24, 772.
46. S. Huang, L.-Z. Cheong, S. Wang, D. Wang and C. Shen, Appl. Surf. Sci., 2018, 452, 67–74.
47. C. Sole, N. E. Drewett and L. J. Hardwick, Faraday Discuss., 2014, 172, 223–237.
48. J.-T. Li, S.-R. Chen, F.-S. Ke, G.-Z. Wei, L. Huang and S.-G. Sun, J. Electroanal. Chem., 2010, 649, 171–176.
49. Y. Wang, Z. Cao, Z. Ma, G. Liu, H. Cheng, Y. Zou, L. Cavallo, Q. Li and J. Ming, ACS Energy Lett., 2023, 8, 1477–1484.
50. M. Qin, Z. Zeng, Q. Wu, H. Yan, M. Liu, Y. Wu, H. Zhang, S. Lei, S. Cheng and J. Xie, Energy Environ. Sci., 2023, 16, 546–556.
51. J. Xu, J. Zhang, T. P. Pollard, Q. Li, S. Tan, S. Hou, H. Wan, F. Chen, H. He, E. Hu, K. Xu, X. Q. Yang, O. Borodin and C. Wang, Nature, 2023, 614, 694–700.
52. N. Yao, X. Chen, X. Shen, R. Zhang, Z.-H. Fu, X.-X. Ma, X.-Q. Zhang, B.-Q. Li and Q. Zhang, Angew. Chem., Int. Ed., 2021, 60, 21473–21478.
53. S. Lei, Z. Zeng, H. Yan, M. Qin, M. Liu, Y. Wu, H. Zhang, S. Cheng and J. Xie, Adv. Funct. Mater., 2023, 33, 2301028.
54. Z. Li, Y. Chen, X. Yun, P. Gao, C. Zheng and P. Xiao, Adv. Funct. Mater., 2023, 33, 2300502.
55. X. Chen and Q. Zhang, Acc. Chem. Res., 2020, 53, 1992–2002.
56. Y. C. Gao, N. Yao, X. Chen, L. Yu, R. Zhang and Q. Zhang, J. Am. Chem. Soc., 2023, 145(43), 23764–23770.
57. A. Maali, T. Cohen-Bouhacina, G. Couturier and J. P. Aime, Phys. Rev. Lett., 2006, 96, 086105.
58. V. Agmo Hernández, ChemTexts, 2023, 9, 10.
59. E. Bonaccurso, M. Kappl and H. Butt, Curr. Opin. Colloid Interface Sci., 2008, 13, 107–119.
60. S. Guriyanova, V. G. Mairanovsky and E. Bonaccurso, J. Colloid Interface Sci., 2011, 360, 800–804.
61. Y. Suzuki, N. Kunikata, M. Kasuya, H. Maki, M. Matsui, K. Kurihara and M. Mizuhata, J. Phys. Chem. C, 2022, 126, 11810–11821.
62. J. Chen, X. Fan, Q. Li, H. Yang, M. R. Khoshi, Y. Xu, S. Hwang, L. Chen, X. Ji, C. Yang, H. He, C. Wang, E. Garfunkel, D. Su, O. Borodin and C. Wang, Nat. Energy, 2020, 5, 386–397.
63. J. Liu, W. Hao, M. Fang, X. Chen, Y. Dong, Y. Chen, Z. Wang, X. Yue and Z. Liang, Nat. Commun., 2024, 15, 9356.
64. A. von Wald Cresce, M. Gobet, O. Borodin, J. Peng, S. M. Russell, E. Wikner, A. Fu, L. Hu, H.-S. Lee, Z. Zhang, X.-Q. Yang, S. Greenbaum, K. Amine and K. Xu, J. Phys. Chem. C, 2015, 119, 27255–27264.
65. L.-L. Jiang, C. Yan, Y.-X. Yao, Y. Lu, J.-Q. Huang and Q. Zhang, Angew. Chem., Int. Ed., 2021, 60(7), 3402–3406.
66. J. F. Ding, R. Xu, N. Yao, X. Chen, Y. Xiao, Y. X. Yao, C. Yan, J. Xie and J. Q. Huang, Angew. Chem., Int. Ed., 2021, 60(20), 11442–11447.
67. W. Yu, K.-Y. Lin, D. T. Boyle, M. T. Tang, Y. Cui, Y. Chen, Z. Yu, R. Xu, Y. Lin, G. Feng, Z. Huang, L. Michalek, W. Li, S. J. Harris, J.-C. Jiang, F. Abild-Pedersen, J. Qin, Y. Cui and Z. Bao, Nat. Chem., 2025, 17, 246–255.
68. G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter Mater. Phys., 1993, 47, 558–561.
69. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, G. A. Petersson, H. Nakatsuji, X. Li, M. Caricato, A. V. Marenich, J. Bloino, B. G. Janesko, R. Gomperts, B. Mennucci, H. P. Hratchian, J. V. Ortiz, A. F. Izmaylov, J. L. Sonnenberg, D. Williams-Young, F. Ding, F. Lipparini, F. Egidi, J. Goings, B. Peng, A. Petrone, T. Henderson, D. Ranasinghe, V. G. Zakrzewski, J. Gao, N. Rega, G. Zheng, W. Liang, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven, K. Throssell, J. A. Montgomery, Jr., J. E. Peralta, F. Ogliaro, M. J. Bearpark, J. J. Heyd, E. N. Brothers, K. N. Kudin, V. N. Staroverov, T. A. Keith, R. Kobayashi, J. Normand, K. Raghavachari, A. P. Rendell, J. C. Burant, S. S. Iyengar, J. Tomasi, M. Cossi, J. M. Millam, M. Klene, C. Adamo, R. Cammi, J. W. Ochterski, R. L. Martin, K. Morokuma, O. Farkas, J. B. Foresman and D. J. Fox, Gaussian 16, Revision C.01, Gaussian, Inc., Wallingford CT, 2016.
70. Q. Zhang and A. Asthagiri, Catal. Today, 2019, 323, 35–43.
71. K. Mathew, R. Sundararaman, K. Letchworth-Weaver, T. A. Arias and R. G. Hennig, J. Chem. Phys., 2014, 140, 084106.
72. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
73. P. J. Stephens, F. J. Devlin, C. F. Chabalowski and M. J. Frisch, J. Phys. Chem., 1994, 98, 11623–11627.
74. S. Grimme, J. Antony, S. Ehrlich and H. Krieg, J. Chem. Phys., 2010, 132, 154104.
75. M. J. Abraham, T. Murtola, R. Schulz, S. Páll, J. C. Smith, B. Hess and E. Lindahl, SoftwareX, 2015, 1, 19–25.
76. W. L. Jorgensen, D. S. Maxwell and J. Tirado-Rives, J. Am. Chem. Soc., 1996, 118, 11225–11236.
77. W. L. Jorgensen and J. Tirado-Rives, Proc. Natl. Acad. Sci. U. S. A., 2005, 102, 6665–6670.
78. A. K. Malde, L. Zuo, M. Breeze, M. Stroet, D. Poger, P. C. Nair, C. Oostenbrink and A. E. Mark, J. Chem. Theory Comput., 2011, 7, 4026–4037.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5ee01076d

‡ Y. C., W. X. and W. Z. contributes equally.

---