Introduction

Given the highest theoretical capacity and natural abundance as the 2nd element in Earth’s crust1, Silicon (Si) -based negative electrodes emerge as promising candidates for lithium-ion batteries (LIBs), addressing the stringent energy demand of today’s electronic devices and electric vehicles2,3,4. However, the commercial application of Si encounters significant hurdles, primarily attributed to the substantial volume expansion and contraction (300%) during cycling. These volumetric fluctuations generate internal stress, which poses a critical threat in all-solid-state lithium-ion batteries (ASSLIBs). Such stress can lead to mechanical degradation of the solid-state electrolyte, compromising the electrode-electrolyte interface and potentially inducing lithium dendrite formation5. Even the most robust solid-state electrolytes (e.g., Li6La3 ZrTaO12, with a hardness of 7.5 GPa) fail to endure the stress imposed by silicon’s volume changes6,7. To solve the inherent challenges associated with Si-based negative electrodes, dissipating internal energy or compensating for volume exchange could be promising alternatives8,9, but accurately and in-situ monitoring the stress at the electrode-electrolyte interface is a prerequisite10,11. Previous studies focused on characterizing internal stress in LIBs have primarily relied on in-situ observation of electrode material volume changes10,11,12,13,14. Techniques such as high-resolution transmission electron microscopy and focus-ion-beam scanning electron microscopy only allow for observing the expansion or contraction of individual nanoparticles15,16. These methods only offer indirect information on internal stress and are confined to a nanoscale detection area, limiting their ability to capture the overall stress distribution within the cell. Moreover, internal stress evolves in real-time during lithium insertion and extraction, intricately linked to the electrochemical reaction process, thereby introducing a more complex mechanical-electrochemical coupling (MEC) effect in ASSLIBs9,13,17,18. Consequently, comprehending the relationship between the evolution of internal stress and electrochemical redox reactions is theoretically important to design Stress-Neutralized batteries.

Current attempts to decipher the MEC effect in LIBs predominantly rely on established commercial electrode materials (LiFePO419, LiCoO220, LiNi0.33Co0.33Mn0.33O221, Li4Ti5O1222, etc.). For example, capacity derivative differential pressure signals (dP/dQ), corresponding to differential volume changes of electrode materials, have proven instrumental in accurately estimating state-of-charge for graphite electrodes. This estimation is crucial for adjusting capacity ratios between negative and positive electrodes (N/P ratios)23. Moreover, a stress balance concept has been realized in ASSLIBs by using LiNi0.8Co0.1Mn0.1O2 (NCM) and LiCoO2 (LCO) as two active materials in the composite positive electrode. This balance arises from the opposite volume changes of NCM and LCO during the charging and discharging processes, effectively canceling out the stress in the composite positive electrode during cycling18. Exploring the next generation of electrode materials, the stack-pressure evolution of Si negative electrodes was investigated in full cells with LiNi0.33Co0.33Mn0.33O2 (NMC111) positive electrodes, revealing an extremely large internal stress (1~2 MPa) due to the substantial volume expansion of the alloy negative electrodes during lithiation/delithiation24. These pioneering studies underscore the significance of understanding the evolution of internal stress during electrochemical processes, considering both the positive and negative electrodes23,25,26,27,28,29,30. However, optimizing the stress balance of one electrode alone is insufficient. Several crucial questions remain unanswered. Primarily, although the correlation among internal stress evolution and electrochemical performance for Si-based negative electrode was established, the strategy to balance such stress are highly desired. Efforts to optimize stress management in silicon-based systems have largely focused on a single electrode, leaving significant gaps in understanding the synergistic effects between negative and positive electrodes’ stresses. For instance, in all-solid-state lithium-metal batteries (ASSLMBs), the molar volume change of lithium metal (13.00 cm3 mol−1) during cycling is often overlooked. The density disparity between lithiated silicon (48.45 cm3 mol−1) and lithium metal exacerbates uniaxial volume changes, leading to overall cell shrinkage during discharge rather than a simple unilateral expansion of Si. This imbalance underscores the need for a holistic strategy that addresses stress evolution across both electrodes in ASSBs. Consequently, there is a 30% shrinkage of the overall volume of the half-cell during discharge, rather than a unilateral expansion of the Si, which exists in all-solid-state lithium-metal batteries.

While sulfur (S)-based positive electrodes exhibit an opposite volume change during cycling compared with Si, offering a natural stress-compensation mechanism, practical realization of this synergy remains limited due to stress heterogeneity and insufficient understanding of internal stress dynamics at the cell level. In earlier studies of liquid batteries, the effects of internal stress in Si–S systems received limited attention because liquid electrolytes can instantly accommodate volume changes of the electrodes without stress buildup. The research efforts predominantly focused on addressing the shuttle effect caused by the dissolution of polysulfides in liquid electrolyte. Some studies have shown strategies in solving the issue, such as improving the chemical compatibility by S@PAN in liquid electrolyte31,32,33 or designing the 3D sulfur copolymer positive electrode34. However, due to the complexity of material preparation and the difficulty in fully eliminating the shuttle effect in liquid electrolyte, there are challenges of practical application. Solid electrolytes promise to physically eliminate the shuttle effect associated with polysulfides. There is a more attractive option for Si–S battery development. The use of gel polymer electrolyte or solid-state polymer electrolyte combined with liquid electrolyte infiltration enables the construction of quasi-solid-state Si–Li2S batteries35,36,37,38. Additionally, the research has successfully assembled Li3.75Si–S cell by adopting polytetrafluoroethylene binders in the dry electrodes39. However, the research studies of Si–S ASSLIBs including the configuration and design are still in the infancy at present. Especially, when switching to an ASSLIBs, the severe volume expansion, which could lead to significant stresses induced by local microstructural changes in a confined space, how these stresses are relieved are integral to battery lifespan. Successful design of Si–S ASSLIBs will be integrally linked to how effectively control the evolution of stress.

Herein, we have developed a solid-state electrochemical cell with a high-precision force sensor for in-situ monitoring of internal stress caused by electrode material volume changes (Fig. 1a and Supplementary Fig. 1). This monitoring, combined with 3D microstructural analysis using high-resolution X-ray micro-computed tomography (micro-CT), enables detailed characterization of stress evolution and its relationship with electrochemical processes. While the opposing volume changes of silicon and sulfur electrodes offer inherent stress compensation, our dynamic volume compensation strategy optimizes this effect by balancing electrode kinetics and addressing stress heterogeneity. These refinements are critical to achieving the demonstrated performance improvements. Moreover, the stress-neutralized 1.24 Ah cell with a high positive electrode active material loading (21.7 mg cm−2) based on commercial silicon and sulfur materials delivered a significantly improved areal capacity of 35.1 mAh cm−2.

Fig. 1: In-situ stress monitoring and Electrochemical performances of Li–Si half-cell and Si–S full cell.
figure 1

a Schematic of the cell assembly and a dynamic volume compensation strategy to design a Si-S full cell in a state of Stress-Neutralization. The internal stress change of b Li–Si half-cell and c Si–S full-cell during cycling, the voltage profiles in the top panel, the stress change in the bottom panel. d, e The internal stress distribution of 20~30 mAh cm−2 Li–Si half-cell and Si–S full-cell.

Full size image

Results

Internal stress regulation

In solid-state batteries (SSBs), unlike liquid batteries, the volume change of electrodes plays a crucial role, as it is closely associated with internal stress changes during the de/lithiation process. These internal stress changes can lead to mechanical degradation of the solid electrolyte, increased interfacial resistance due to delamination and voids formation, ultimately affecting the cells’ stability. In order to elucidate the influence of volume change-induced internal stress, we assembled a stress-unbalanced Li–Si half-cell and a stress-neutralized Si–S full cell, both having the same theoretical capacity and negative/positive capacity ratio (N/P ratio) of 1.1. The stack stress generated by the volume expansion of the sulfur positive electrode could dynamically volume buffer the stress resulting from the volume contraction of the Si negative electrode during discharge. Figure 1b, c present galvanostatic cycling curves with correlated stress measurements from the Li–Si half-cell and Si–S full cell at a current density of 2.0 mA cm−2. In-situ pressure monitoring demonstrates that while the natural stress-compensation effect reduces overall stress evolution, localized stress peaks persist, such as in the Li–Si system (Fig. 1b). These peaks are attributed to mismatched electrode kinetics and nonuniform volume change. By incorporating dynamic structural adjustments, such as electrode design and density matching, our strategy minimizes these stress variations, achieving a nearly uniform stress distribution. Figure 1b, c show a comparative analysis of stress evolution in systems with and without dynamic compensation design, highlighting a ~93.6% reduction in stress heterogeneity in the optimized cell. Despite a significant capacity loss (55%) in half-cell due to stress variation in the Li–Si half-cell contributing to poor electrolyte-electrode interfacial contact, the stress reached as high as −8.53 MPa after discharge. Alternatively, the Si–S full cell with a mass loading of 12 mg sulfur cm−2, where the S positive electrode was found to deliver reversible capacities of 1558.7 mAh g−1 (14.68 mAh) with 93% utilization efficiency. The stress change range observed in the Si–S full cell is only from 0 to −2.86 MPa throughout the entire process (Fig. 1c).

When the active silicon loading in the half-cell is too high, on one hand, the increase in current density leads to abnormal lithium deposition and consequently short-circuiting, and on the other hand, the unmitigated volume expansion results in a sharp increase in stress, leading to capacity degradation. Thus, we simulated the stress data after complete conversion of the active material. According to the internal stress distribution analyzed through finite element modeling performed using ABAQUS, we simulated the internal stress distribution under varying areal capacities (10~30 mAh cm−2) in Fig. 1d, e and Supplementary Fig. 2a–d. With increasing areal capacity, the thickness of the electrode layer increases, and the amount of Li⁺ migration during the reaction also grows, resulting in a higher overall internal stress within the battery. However, the half-cells exhibit similar stress concentration regions under different areal capacities. For instance, in the case of 30 mAh cm−2 half-cell, the significant volume expansion undoubtedly leads to high pressure levels (24~48 MPa, Fig. 1d). The stress generated by the expansion and contraction of the electrode is transmitted to the solid-state electrolyte (SSE) layer. The maximum stress in the SSE was found at the edge with a value of 35.97 MPa. This high stress concentration at the SSE edge can lead to mechanical damage, such as cracks and fracture, which damages structural integrity and ionic conductivity of SSE. The alternation of electrode component to the full-cell configuration is calculated through finite element modeling (Fig. 1e). The maximum internal stress value declined to 4.48 MPa on SSE which distributed along the interface, it mitigated the risk of structure damage due to high stress concentration on SSE and maintained the structural stability of the interface. Supplementary Fig. 2e illustrates the experimental and simulated external pressure values of half-cell and full-cell systems under different areal capacity. The significant volume changes of electrodes during the charge and discharge processes in the half-cell cannot be effectively compensated, leading to a rapid increase in the pressure exerted on the electrolyte. In contrast, the holistic design of the full-cell effectively offsets the volume changes of individual electrodes, achieving self-elimination of stresses.

Dynamic volume compensation principle

In order to further verify the dynamic volume compensation principle, operando scanning via in-situ X-ray micro-CT at a voxel size of 6.65 µm was conducted during the charge/discharge process. Figure 2a illustrates three X-ray tomograms corresponding to different states of the Li–Si half-cell: 100% state of charge (SOC), 100% depth of discharge (DOD), and 100% SOC. After lithiation, the Si electrode’s expansion increased from 163.6 to 196.5 µm, while the lithium negative electrode thickness decreased by 50.5 µm, the significant volume changes in the electrodes resulted in shifting at the electrode-electrolyte interface, generating substantial internal stress during the process. Subsequently, the overall volume of the half-cell decreased by 2.76%, with the thickness returning to the pristine state after delithiation of the Li22Si5 alloy. Rather than attempting to prevent drastic volume changes, a more effective approach involves managing and mitigating these fluctuations. Figure 2b compares a Si–S full cell with the half-cell (Fig. 2a and Supplementary Figs. 3 and 4), the thickness of the full cell experiences minimal change, reducing by only 1.3 µm (0.21%) from lithiation to delithiation throughout the entire process. When the half-cell discharges, the thickness of the silicon electrode expands by 32.9 micrometers, while the thickness of the lithium negative electrode decreases by 50.5 micrometers compared to the original. Importantly, a substantial volume changes (13.5%, calculated from Fig. 2c and Supplementary Table 1) in the entire system resulted in high stress on the electrolytes. Moreover, the concentrated force from the lithium plating/stripping could easily induce plastic deformation, even cause cracks in brittle materials like LiBH4. In order to further verify the dynamic volume compensation principle, high-resolution X-ray micro-CT provides a three-dimensional view of electrode volume changes. As shown in Fig. 2d and Supplementary Fig. 5, the Li–Si system results in uneven volume adaptation, with observable microstructural cracks near the electrode-electrolyte interface. In contrast, the dynamic volume compensation strategy in Si–S system mitigates these distortions, preserving electrode integrity over extended cycles. This is reflected in the superior cycling stability and capacity retention observed in the following measurements of Si–S full cell. As illustrated in Fig. 2f, and Supplementary Figs. 6, 7, electrode cracking caused by large internal stress or nonuniform deposition generally occurs in various electrode types, including lithium metal negative electrode (Li), conversion type electrode (S, et al.), alloy type electrode (Si, et al.), and common LiFePO4 (LFP) electrode. The active material particles in the electrode experienced inter- and intra-particle stress during the swelling and shrinkage process, leading to the fracture and pulverization of the particles. This resulted in the inactivation of active material in subsequent cycles. Subsequently, multiple results show that the battery experiences short circuits, capacity degradations, and decreases in Coulombic efficiency (Supplementary Figs. 8–10). For the discharge of Si–S full cell, the thickness of the sulfur (S) electrode expands by 30.5 micrometers, while the thickness of the Li22Si5 negative electrode decreases by 31.8 micrometers compared to the original. Consequently, the volume change is a mere 0.79 % (Fig. 2c). The volume changes of the negative and positive electrodes in the battery mutually compensated, maintaining stress balance, and achieving stability in the overall system. This controlled rate ensures the stability of the interface structure and the integrity of electrode, guaranteeing electrochemical stability during cycling (Fig. 2e, g and Supplementary Fig. 11). It is noteworthy that this stress-neutralization strategy is not limited to the Si–S full cell and is also applicable to Si–Se ASSLIBs (Supplementary Fig. 12).

Fig. 2: Dynamic volume change of Li–Si half-cell and Si–S full cell.
figure 2

a, b Operando X-ray CT images of the Li–Si half-cell and Si–S full cell. c Volume changes for the active materials including the positive and the negative electrodes at 100% SOC and DOD are imported for comparison with the rate of volume change. d, e The 3D volume rendering of Li–Si half-cell and Si–S full cell. f SEM images of positive electrodes before and after cycling in half-cells (Li–Si, and Li–S). g SEM images of electrodes before and after cycling in Si–S full-cell.

Full size image

Stress-neutralized Si–S full cell

To examine the effectiveness of stress-neutralization strategy in ASSLIBs, we further increased the sulfur loadings to 42.6 mg cm−2 in Fig. 3a. Notably, the achieved areal capacity surpasses 71.5 mAh cm−2, significantly exceeding the benchmark value of 3 mAh cm−2 set for commercial LIBs. This accomplishment results in a reliable specific energy of 525 Wh kg−1 based on total cell mass (Supplementary Tables 2 and 3). To achieve a significantly high energy density, thick positive electrode architectures with a low solid-state electrolyte fraction are essential. In Supplementary Fig. 13, X-ray tomograms of a high-capacity single cell reveal a positive electrode thickness of up to 1 mm. Notably, the maximum loading capacity that a single cell can withstand was tested based on the results of thick electrodes. Rate performance (Fig. 3b, c) at a high positive electrode active material loading (18.1 mg cm−2) exhibited specific discharge capacities of 1651, 1512, 1348, 1254, and 759 mAh g−1 (corresponding to large reversible areal capacities of 30.1, 27.3, 24.3, 22.6, and 13.6 mAh cm−2) at various current densities of 2.51, 5.02, 10.05, 12.56, and 25.12 mA cm−2, respectively. Even at a high current density of 25.12 mA cm−2, the electrolyte did not short-circuit, validating the dynamic volume compensation strategy’s ability to maintain battery stability through internal stress neutralization. Long-term cycling of the full cell (mass loading of 21.0 mg sulfur cm−2) demonstrated a large areal capacity over 34.9 mAh cm−2 (1661 mAh g−1) even at a high current density of 3.5 mA cm−2. The average Coulombic Efficiency (CE) remained 100.01% (standard deviation was 0.22%) from 3rd to 500th, showcasing an impressive areal capacity of 19.0 mAh cm−2 (Fig. 3d). Notably, this stability outperforms typical Li–S batteries, which are known for comparatively lower stability than LIBs.

Fig. 3: Electrochemical performances of stress-neutralized Si–S full cell.
figure 3

a High energy density performance, b and c Rate performance with a high S loading of 18.1 mg cm−2, d Cycling stability test of stress-neutralized Si–S full cell.

Full size image

Kinetic processes analysis

To investigate relationship between stress evolution and reactions kinetics, we performed kinetic analyses using Electrochemical Impedance Spectroscopy (EIS) coupled with Distribution of Relaxation Times (DRT). The DRT method offers significant advantages over traditional circuit-fitting approaches by allowing the decomposition of impedance spectra into distinct time constants, thereby isolating individual electrochemical processes with reduced subjective fitting errors40,41. This is particularly important for all-solid-state batteries (ASSBs), where complex overlapping processes often obscure Nyquist plots.

Li–S, Si–S, and Li–LiSi cells have been assembled, respectively. DRT analyses of Li–S and Si–S cells revealed distinct differences in their impedance profiles under different states of charge (SOC) (Fig. 4). Peaks in the DRT profiles at 10−4~10−2 s correspond to the Li⁺ transport across the cathode-electrolyte interphase (Rcei), while peaks at 10−2~100 s represent charge transfer at the positive electrode surface (Rct). These combined resistances (Rint = Rcei + Rct)42,43,44 dominate interfacial impedance, while peaks at lower frequencies (around 0.1 Hz) represent solid-state diffusion of Li⁺ in electrodes (Rd), which is driven by the concentration. We conducted EIS measurements of Li–S and Si–S at different SOC in the third cycle for DRT analysis. At 100% SOC, Si–S exhibited significantly lower overall impedance compared to Li–S (Fig. 4a). The Rint of Si–S is less compared to Li–S from DRT data. The significant difference is derived from the evolution of the Li+ transport kinetics at the positive electrode/SSE interface. This observation suggests that Si–S exhibits a better interfacial stability and maintains the close contact interface between positive electrode and SSE after several cycles, facilitating better ionic and electronic conduction compared to Li–S (Fig. 4b, c).

Fig. 4: DRT analysis of Li–S and Si–S cells’ kinetic processes.
figure 4

a EIS spectra of the Li–S half-cell, Si–S full-cell at 100% SOC, b, c Corresponding DRT profiles transformation of EIS. d EIS spectra of the Li–S half-cell at 80~20% SOC, e corresponding DRT profiles, and f 2D intensity color map. g EIS spectra of the Si–S full-cell at 80~20% SOC, h Corresponding DRT profiles, and i 2D intensity color map.

Full size image

Figure 4d, g represent the EIS data of Li–S and Si–S cells at different SOC (80~20%), respectively. While corresponding DRT profiles are shown in Fig. 4e, f, h, i. Rd peak intensity of Li–S and Si–S decrease significantly from 100% to 80% SOC. The reduction of impendence is attributed to the migration of Li+ during the initial discharge stage, creating a concentration gradient within the S positive electrode to improve the diffusion kinetics. As discharge continues, it can be observed a sharp increase of Rd peak intensity from 60% to 20% SOC in Fig. 4e, f, reflecting hindered solid diffusion of Li+ in the positive electrode due to poor interfacial contact and stress-induced degradation. Conversely, the intensity of Rd peak gradually increases when SOC is from 60% to 20% in the Si–S cell, indicating stable diffusion kinetics through the discharge process (Fig. 4h, i). Meanwhile, Rint of the Li–S cell shows marked peaks during discharge in Fig. 4f, attributed to large volume changes causing voids and poor interfacial contact. In contrast, the Si–S cell reveals consistent lower Rint than Li–S, and maintains a stable interface (Fig. 4i). The microscopic changes of resistance in the above results reflect the macroscopic interface state. Si–S system maintains a stability of interface, attributed to closer interfacial connections facilitated by Dynamic Volume Compensation. The Li–S system is affected by the huge volume change of the Li metal negative electrode, which makes it impossible to realize the spontaneous Dynamic Volume Compensation, showing an unstable state interface. The discharge process in the Si–S system reflects the combination of Li–S discharge and Li–LiSi charge. During discharge, the Li+ extraction rate from the LiSi negative electrode (volume contraction) exceeded the rate of Li+ insertion into the sulfur positive electrode (volume expansion) in Fig. 4e and Supplementary Fig. 14b–d. This resulted in a gradual stress decrease, which was eventually balanced as Li2S formation completed. This improved kinetics translated into better performance, with Si–S achieving a capacity of 1736 mAh g−1, significantly higher than the Li–S cell’s 1263 mAh g−1 (Supplementary Fig. 14a). The stress-neutralized Si–S system demonstrated stable interfacial ionic conductivity and suppressed stress changes, enabling higher capacity release and longer cycle stability. These results underscore the importance of stress management in improving reaction kinetics and mitigating performance degradation in ASSBs. The coupling of stress evolution with electrochemical behavior highlights the role of Dynamic Volume Compensation in buffering stress changes, enhancing Li+ transport, and stabilizing interfaces in the Si–S system. This study demonstrates that addressing mechanical-electrochemical interfacial challenges is pivotal for advancing high-performance ASSBs.

Wide temperature, cell pack and large prototype (Ah sample)

We further illustrate the cell performance and highlight the transition from fundamental research to practical applications based on stress-neutralization strategy. Firstly, the operation across a wide temperature range demonstrates the versatility and safety of ASSLIBs in potential future applications. As shown in Fig. 5a, the cell exhibits high specific discharge capacities at various temperatures: 1673 mAh g−1 (10th), 1584 mAh g−1 (15th), 1456 mAh g−1 (20th), 1276 mAh g−1 (25th), and 1135 mAh g−1 (30th) at 120, 100, 80, 60, and 37 °C, respectively. Remarkably, even at 37 °C, the battery maintained a reversible capacity of nearly 70%. Simultaneously, the adoption of a series/parallel pack design proves advantageous for enhancing output voltage and reducing inactive material components at the package and module levels, making the technology more viable for practical applications (Fig. 5b, c). Supplementary Fig. 15 illustrates a series and a parallel pack cell, each composed of two single cells, achieving a same areal discharge capacity of 36.9 mAh cm−2 at 1.25 mA cm−2. Whether in series or parallel, the battery packs have exhibited extremely high Coulombic efficiency and cycling stability. As shown in Fig. 5d, an Ah-level cell was assembled and evaluated to demonstrate the cycling stability of large prototype. Under a high positive electrode active material loading (21.7 mg cm−2), the cell exhibited low voltage polarization, and the capacity reaches 1.24 Ah (1622 mAh g−1) at 0.05 C. Subsequently, the cell had a high retention of capacity (0.82 Ah) after 100 cycles at 0.1 C (Fig. 5e).

Fig. 5: Wide temperature performance, cell pack and Ah-level sample of Si–S cells.
figure 5

a The electrochemical performance of Si–S full cell at a wide temperature range, b The voltage profiles of the series double cells, c The voltage profiles of the parallel double cell, d The voltage profiles of Ah-level cell at 0.05C, e Cycling stability of the Ah-level cell.

Full size image

Discussion

The mechanical-electrochemical findings presented in this study underscore the efficacy of the dynamic volume compensation strategy in achieving stress-neutralization within all-solid-state Si-S full cells. By leveraging and optimizing the inherent opposing volume changes of silicon and sulfur electrodes, this strategy addresses critical challenges associated with stress heterogeneity, electrode integrity, and interfacial stability during cycling. Notably, the 1.24 Ah cell achieved a notably improved areal capacity of 35.1 mAh cm−2, supported by a high positive electrode mass loading of 765 mg of active sulfur, a low N/P ratio, and robust performance across a wide range of current densities and temperatures. These results highlight the practical implementation of our design, including high-precision stress monitoring, structural optimization, and advanced 3D microstructural analysis, which collectively enable the superior cycling stability and energy density demonstrated. In summary, the dynamic volume compensation strategy provides a scalable and effective approach to overcoming the mechanical-electrochemical challenges inherent in alloy- and conversion-type materials. This innovation not only accelerates the development of high-performance silicon-sulfur batteries but also provides a framework for advancing next-generation post-Li-ion energy storage technologies.

Methods

Cell fabrication

Half-cell with different positive electrode materials were assembled. All electrode materials were prepared using 4 mm ZrO₂ milling balls in an argon atmosphere, without heating. For Si positive electrode composite, Si nano-powders (≥99.9%, Alfa Aesar, Supplementary Figs. 16 and 17) was mixed with active carbon (Japan LION) and LiBH4 (95%, Sigma-Aldrich) by simple ball milling (Fritsch P7) under 150 rpm for 1 h, with an intermittent 30 min pause. For the S (99.98%, Sigma-Aldrich), Se (≥99.5%, Sigma-Aldrich) positive electrode, the active material was initially mixed with activated carbon at 400 rpm for 3 h, and the mixture was then combined with the Li6PS5Cl solid-state electrolyte (≥99.5%, China Automotive Battery Research Institute Co., Ltd.) by ball milling at 150 rpm for an additional hour. Commercial electrode was used directly in LiFePO4 half-cell. Li foil (99.95%, China Energy Lithium Co., Ltd., Tianjin) was compressed (RIKEN KIKI CO., Ltd) to around 100 µm (diameter = 10 mm) as negative electrode under the pressure of 20000 N in an argon-filled glove box (H2O < 0.01 ppm, O2 < 0.01 ppm). For full-cell, Li22Si5 alloy (Supplementary Fig. 17) was ball milled with SSE and active carbon at 150 rpm for 1 h as negative electrode. And the electrochemical performances of prepared Li22Si5 alloy negative electrodes are shown in Supplementary Fig. 3. S composite material was used as positive electrode. For full cells, the N/P ration was set as 1.1 according to this calculation. The results of lower N/P ratios are shown in Supplementary Fig. 18.

For all cells, LiBH4, Li4(BH4)3I, Li6PS5Cl and Li3InCl6 powder were used as electrolyte. The batteries were assembled and pressed in polyether ether ketone (PEEK) dies (diameter = 10 mm) under the external stress of 40 MPa. The results of Si–S cells used other different electrolytes (Li4(BH4)3I, Li6PS5Cl and Li3InCl6) are in Supplementary Figs. 19–21. The cells parameters in manuscript are in Supplementary Table 4.

Stress and electrochemistry measurements

The stress measurements were conducted using a force sensor (Guangzhou SimBaTouch Electronic Technology Co., Ltd.), rated for 9.8 kN, which was positioned at the bottom of the cell (Fig. 1). Before starting the stress tests, the force sensor was calibrated with weights to ensure accuracy. Following calibration, the battery mold, force sensor, and fixture were assembled together, applying pressure to the battery through bolts on the fixture. The force sensor was connected to a data collector on computer, enabling real-time data visualization. After applying pressure, all cells were held at open circuit over 15 h before cycling to reduce the stress drift after applying the initial external pressure of 40 MPa. Only when the stress readings on the monitoring software indicated stability were further tests conducted. A photograph of the real cell setup components designed by Ningbo Chuangli New Material Technology Co., Ltd. is in Supplementary Fig. 1. The electrochemical measurements of the cells were used a LAND CT3001A (Wuhan LAND electronics Co. Ltd., China) battery test system. Voltage windows were chosen to be appropriate for the different cells tested. Li–Si half-cells and Li–Li22Si5 alloy half-cell were cycled between 0.01 and 1 V. The voltage window of Li–S and Li–Se half-cell was between 1 and 2.5 V. And that of Li–LiFePO4 was 2.5~4.0 V. For Si–S full-cells, the voltage windows were 0.1/0.5~3.0 V and 0.2~6 V. Si–Se full-cells are 0.5~3.0 V. The multiple results of Li–Si, Li–S and Li–LiFePO4 cells are shown in Supplementary Figs. 8–10. The cells used LiBH4 and Li4(BH4)3I as electrolyte were measured at 120 °C and 35~120 °C with static heating.

Sample Characterization

Phase identification was measured by using X-ray diffraction (XRD, PANalytical Empyrean) with a copper Kα line (λ = 0.1542 nm). All samples for XRD analysis were prepared in a glovebox filled with pure Ar (O2 < 0.01 ppm, H2O < 0.01 ppm) and sealed with Kapton® films. Scanning Electron Microscope (SEM, Hitachi Regulus 8100) was applied by using an accelerating voltage at 20 kV to observe the morphology of electrode samples before and after cycling. The synchrotron X-ray tomography measurements were tested on Zeiss Xradia 620 Versa with a voxel size of 6.65 µm. The cells were rotated by 180° with respect to the normal axis of the detector and around 1000 slices of 2D projections were taken. The data analysis and images were generated using Dragonfly software, Version Dragonfly Pro for Windows. Comet Technologies Canada Inc., Montreal, Canada. Electrochemical impedance spectroscopy (EIS) measurements were implemented using a Bio-Logic SP-300 (France). The input frequency range was set to 4 MHz to 0.05 Hz (6 data points per decade of frequency) with an amplitude of 10 mV under the stable open circuit voltage condition.

ABAQUS simulation

The finite element modeling was conducted to analyze the internal stress distribution in Li–Si half-cell and Si–S full cell using ABAQUS. A sandwich structure composed of negative electrode (Li, LiSi alloy), SSE (LiBH4), and the positive electrode (Si, S) was constructed, with the external pressure set according to our operando X-ray CT results. The material parameters including their density, Young’s modulus and Possion’s ratio were listed in Supplementary Table 1.