Revealing the dynamic evolution of Li filaments within solid electrolytes by operando small-angle neutron scattering

doi:10.21203/rs.3.rs-1520454/v1

Download PDF

Article

Revealing the dynamic evolution of Li filaments within solid electrolytes by operando small-angle neutron scattering

https://doi.org/10.21203/rs.3.rs-1520454/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Lithium dendrites (filaments) propagation in the solid electrolyte (SEs) leading to short circuits is one of the biggest obstacles to the application of all-solid-state lithium metal batteries (ASSLMBs). Due to the lacking of operando techniques that can provide nano-scale spatial resolution, the insufficient knowledge of the lithium dendrite growth inside SEs makes it difficult to suppress dendrite formation. To reveal the mechanism of Li filament growth in SEs, we achieved real-time monitoring of the nanoscale Li filament growth by operando small-angle neutron scattering (SANS) in a Cu|LLZNO|Li cell. On continuous plating, Li filament growth is not simply an accumulation of Li, but there is a dynamic evolution due to the competition between Li filament formation and self-healing. This dynamic evolution of Li filaments has been proved to be prevalent in typical SEs. With the aid of simulations and experiments, this dynamic competition was demonstrated to be highly dependent on the temperature variation. The enhanced self-healing ability of Li at elevated temperatures plays a positive role in suppressing Li filament formation. The cycle life of the cell was improved by heat therapy, which provided a brand-new insight for suppressing Li filament formation. Operando SANS with high Li sensitivity and high spatial resolution provides a new platform for investigating Li filaments in SEs.

Lithium metal is regarded as the ultimate anode for high-energy rechargeable batteries, in which the final hurdle is to eliminate the lithium dendrite propagation.^{1, 2} To achieve this goal, solid electrolytes (SEs) with high shear modulus and Li⁺ transference numbers are anticipated to couple with lithium metal.^{3, 4} However, far from the expected result, the critical current density of SEs, at which the cell will be shorted, is much smaller than that of liquid electrolytes.^{5, 6} SEs, orders of magnitude stiffer than lithium, though, fail to suppress the dendrite propagation, demonstrating that SEs can be served as the breeding ground for dendrite instead of the barrier layer.^{7, 8}

The dendrite propagation, leading to catastrophic failure, can be considered as ‘cancer’ in all-solid-state lithium metal batteries (ASSLMBs). Great efforts have been devoted to revealing the propagation mechanism and preventing dendrite growth, which strongly depends on advanced characterization technologies.^9-13 State-of-the-art techniques for tracking the evolution of Li dendrites at different scales are listed in Fig. 1. The electron microscope (SEM/TEM) is expected to provide morphology and structure information from the bulk phase down to the atomic scale. More recently, with the aid of high-resolution TEM under a simulated Li-enriched environment, Chi et al suggested that Li⁺ encounters with electrons at grain boundaries, forming local Li filaments which are more appropriately referred to as the segregated Li inside SEs instead of traditionally defined ‘Li dendrite’.¹⁰ Moreover, only when these filaments form extensive interconnected networks inside the entire SEs, the short circuit would eventually occur and lead to the failure of a cell. It is bold to speculate if these filaments inside grain boundary can be detected at an early stage, the remedial measure can be timely adopted to prolong the life span of ASSLMBs. Thus, the critical step lies in developing an operando and non-destructive characterization technique with high sensitivity to Li element and grain boundary under real electrochemical testing conditions. However, it remains a big challenge to existing characterization techniques, until now there has been no one that can meet all these demands.⁹

Herein, Li-sensitive and spatially diagnostic small-angle neutron scattering (SANS) has been employed to observe the dynamic Li filament growth for the first time. Benefiting from the intrinsic sensitivity towards Li and grain boundaries, this operando SANS confirmed the Li filaments initially incubate within the SEs grain boundary and followed by dynamic competition between Li filament growth and self-healing, which is proved to be universal in typical SEs. With the aid of theoretical simulation and experimental verification, the dynamic evolution of Li filaments was demonstrated to be highly dependent on the temperature variation. The enhanced self-healing ability at elevated temperatures plays a positive role in preventing Li filament formation and even can eliminate the existing Li filaments. The cycle life of ASSLMBs can be improved by heat therapy, which provides new insight into the suppression of Li filaments.

Difficulties in learning the growth of Li filaments in SEs. The buried and localized nature of Li filaments in SEs makes it more complicated and difficult in the corresponding characterization, not to mention operando and nondestructive applications. ^{9, 14} State-of-the-art techniques for tracking the evolution of Li dendrites at different scales are listed in Fig. 1, including optical microscope,^{15, 16} nuclear magnetic resonance (NMR),^{17, 18} in-situ neutron depth profiling (NDP),^{12, 13} synchrotron X-ray imaging,^{11, 19, 20} scanning electron microscope (SEM)^{21, 22} and transmission electron microscope (TEM).^{23, 24} Amongst, optical microscope, NDP, and synchrotron X-ray techniques, as listed in the dark green box of Fig. 1, can be fitted with delicately designed devices for in-situ observations of Li dendrites under working condition. However, the spatial resolution is limited to tens or even hundreds of micrometers, which is too low to observe the initial growing period of Li filaments. To some extent, when Li dendrites are detected by these techniques at the micrometer scale, the short circuit of the cell has already happened or been upcoming. On the other hand, to obtain a high spatial resolution, conventional SEM and TEM analyses are employed to provide a wealth of information down to the atomic scale, yet are limited by the instability of lithium under electron beam irradiation. Moreover, the current in-situ observation is under the simulated Li-enriched environment, not the real electrochemical cell. The SEM and TEM characterizations, in a sense, fail to provide the time-resolved Li filament evolution under real electrochemical testing. To obtain a high spatial resolution, an in-situ transmission grazing-incidence small-angle X-ray scattering (GISAXS) method was employed to investigate the nucleation and growth of Li electrodeposition in a liquid system.²⁵ Unlike X-ray, neutrons, which interact with nuclei rather than electrons, have a higher sensitivity to light elements such as H and Li. Small-angle neutron scattering (SANS), which is similar in principle to SAXS, is better at monitoring Li filament evolution in an all-solid system, due to the higher Li sensitivity and stronger penetrating ability of neutrons than X-ray.^{26, 27}

The combination of high Li sensitivity and high spatial resolution makes operando SANS well suited for a whole-process monitor of the Li filament evolution in SEs. To demonstrate this speculation, Li₆.₅La₃Zr₁.₅Nb₀.₅O₁₂(LLZNO) with a typical Ia-3 structure,²⁸ as exhibited in Supplementary Fig. 1 and Supplementary Table 1,2, was utilized as the SEs to assemble the solid-state cell. As demonstrated in Fig. 2a,b, the Cu|LLZNO|Li cell is immobilized in an independently designed and developed device, which is then mounted in a neutron beamline. The device integrates multiple functions of temperature, atmosphere, and current control. To ensure an effective signal from Li filaments within the SEs, the Cu ring is used as the electrode sheet to eliminate intrusive signals from the Cu ring and the Li metal deposited on it. The finite element method (FEM) was applied to simulate the growth of Li dendrites on a macroscopic scale in this particular cell. (Supplementary Fig. 2) The special Cu ring configuration results in a locally enhanced electric field distribution. The inhomogeneous electric field induces inhomogeneous deposition of Li, similar to the inhomogeneous deposition caused by the "edge effect" around the electrode sheet.^{11, 29} Therefore, the Li dendrites grow diagonally from the edge of the Cu ring toward the counter electrode, which ensures effective detection of the Li dendrite in the neutron detection region. It’s worth noting that the thickness of the LLZNO pellet is 1.0 mm, far beyond the penetrating capability of X-ray (less than 100 µm). However, the neutron transmittance of LLZNO pellet and Li foil were tested to be 68.26% and 88.26%, respectively, which are high enough to meet the requirements of operando SANS experiments. Further tests were performed to compare the two-dimensional SANS spectra of Li foil and LLZNO pellet, as shown in Fig. 2c. The Li foil exhibits almost no small-angle scattering signal. Since SANS is a characterization method for probing inhomogeneities in materials at the nanoscale, with no signal response for homogeneous substances.³⁰ Hence, although a large number of neutrons directly penetrating the Li foil are captured on the probe panel, these count points are disorderly distributed and cannot be transformed into a valid SANS intensity curve by integration. In contrast, the internally inhomogeneous LLZNO pellet exhibits a strong small-angle scattering signal (Fig. 2d). The different response to the SANS signal ensures that the microstructural variation induced by the precipitation of Li filaments in LLZNO pellets can be accurately reflected on the corresponding SANS intensity curve.

Operando monitoring of Li filament growth in SEs. The filament growth inside the LLZNO pellet was investigated under a unidirectional and gradual increasing current density to obtain a rapid Li filament growth and collect valid SANS signals. Theoretically, the higher the current density is, the faster the deposition rate becomes. Once reach the critical current density, the Li filaments grow widely and eventually lead to the dendrite and constitute a short. Here, Li-ions are only plated onto the Cu ring in one direction to exclude any complications from the residual ‘dead lithium’ due to the incomplete removal of Li during stripping. As shown in the voltage profile of Fig. 3a, the polarization potential of Li plating increases with the current increases during the initial stage, in line with Ohm's law. However, as the current density continues to increase, the polarization potential suddenly drops to a very small value, as indicated by the black arrow. This phenomenon can be attributed to the accumulation of Li metal on the Cu ring, increasing the contact area between Cu and LLZNO, and further decreasing the interface impedance (Supplementary Figs. 3,4). Although the polarization potential decreases, the complete short circuit does not occur. Instead, the voltage fluctuates with prolonged time, which is tentatively speculated to be caused by the dynamic evolution of Li filaments inside the SEs.

During the plating of Li, the synchronous neutron scattering intensity variation is recorded and exhibited in Fig. 3b. The relationship between the scattering intensity and the number of scatterers is shown below.³⁰

$$I\left(q\right)=n{V}^{2}\varDelta {\rho }_{SLD}^{2}P\left(q\right)S\left(q\right)+{B}_{inc }$$

Where I(q) is the scattering intensity, q is the scattering variable, n is the number of scatterers (Li filaments in this system), V is the volume of scattering, $\varDelta {\rho }_{SLD}$ is the scattering length density contrast between the Li and SEs, P(q) is the form factor, S(q) is the structure factor, and B_inc is the incoherent scattering background. In this scattering system, the $\varDelta {\rho }_{SLD}$ is a fixed value and other parameters are constant. Therefore, the enhancement of scattering intensity I(q) is caused by the increase of Li filaments inside SEs. To confirm this, we tested the variation of the scattering intensity of LLZNO when no current was applied. As shown in Supplementary Fig. 5, the intermediate value of the relative scattering intensity fluctuates around 1, or even less than 1. In comparison, the red fitted curve in Fig. 3b shows that the intermediate value of the scattering intensity is always greater than 1. The scattering intensity increases at the early stages of Li deposition, indicating the precipitation of Li filaments inside the LLZNO. Consistent with previous reports, this initially increased scattering intensity of the Li filament propagation is due to the abundance of grain boundaries or defect sites in LLZNO SEs.^{10, 11} In general, the growth of Li filaments should be faster as the current density gradually increases, and the scattering intensity should increase sharply. However, scattering intensity becomes fluctuant with further current increment, instead of proleptic increment, demonstrating the Li filaments do not accumulate continuously even upon ongoing deposition. Therefore, it can be speculated that the penetration of Li filaments inside SEs is not a monotonic accumulation, but a complex dynamic evolution.

The SANS intensity curves collected at different current densities were resolved to reveal the evolution of Li filaments at the nanoscale (Fig. 3c). There is no obvious trend in these curves as the current density increases. Similar fluctuations confirm the irregular evolution of Li filaments. Fit these SANS intensity curves by the Guinier-Porod model to quantitatively evaluate the microstructural variation in SEs, especially at grain boundaries:³¹

$$I\left(q\right)=\frac{G}{{q}^{s}} exp\left(\frac{{-q}^{2}{R}_{g}^{2}}{3-s}\right), \text{f}\text{o}\text{r} q\le {q}_{1}$$

$$I\left(q\right)=\frac{D}{{q}^{d}}, \text{f}\text{o}\text{r} q\ge {q}_{1}$$

Where G is the Guinier scale factor, D is the Porod scale factor, s is the dimension variable, R_g is the radius of gyration, and d is the Porod exponent (also known as the Power-law). q₁ can be calculated by:

$${q}_{1}=\frac{1}{{R}_{g}}{\left(\frac{3d}{2}\right)}^{1/2}$$

Since the dimension variable s of this system is a constant 1.0, R_g corresponds to an approximate 2D lamella structure, which could be assigned as Li filaments at grain boundaries. The relationship between R_g and the thickness (T) of the randomly oriented sheet is:

$$T={\sqrt{12}·R}_{g}$$

In this system, T represents the thickness of the grain boundary, that is, the T value will increase when Li filaments grow within grain boundaries. The detailed fitting parameters of the Cu|LLZNO|Li cell are shown in Table S3. The initial T₀ is 5.09 nm before the Li plating process when no filaments are present in grain boundaries (Fig. 3d). At the early stage, this T value gradually increases to 5.43 nm with lifted current density and prolonged time, which is due to the formed Li filaments filling in grain boundaries. The T value increment is terminated at 25 µA cm^− 2 and replaced by fluctuation between 5.09 nm and 5.43 nm during the following progress.

On the other hand, the three-dimensional structures of Li filaments can be investigated by using the fractal analysis on small-angle scattering, due to the correspondence between the fractal dimension and Power-law.^{32, 33} It should be noted that for the Power-law dependence, the range of q should be carefully selected within 2π/D < q < 2π/d, where D is the upper limit for the mass fractal structure and d is the primary unit of the structure. As illustrated in Fig. 3d, the obtained Power-law is between 3 and 4, corresponding to the surface fractal dimension. The roughness of the Li filament surface in this system can be estimated based on the Power-law value. The lower the value is, the rougher the surface is. During the entire Li plating, the Power-law value firstly decreases from 3.81 to 3.72, indicating the rougher surface of Li filaments, which corresponds to the continuous formation of Li filaments at grain boundaries. After 25 µA cm^− 2, a similar fluctuation can be also observed in the Power-law value. Taking both T and Power-law value into account, it can be deduced that at the early stage, Li filaments emerge and expand grain boundaries, as reflected by thicker grain boundaries and rougher Li filament surface. Then the abnormal phenomenon occurs that Li filaments partially disappear rather than continuedly to accumulate even under a higher current density, that is, the self-healing process.^34–38 The Li filament growth and self-healing compete with each other and result in fluctuating parameters of T and Power-law. This mutual competition also leads to huge voltage fluctuations and scattering signal fluctuations.

To investigate the effect of temperature on the dynamic evolution of Li filaments, the same operando SANS test was performed on Cu|LLZNO|Li at 80℃ (Supplementary Fig. 6 and Supplementary Table 4). The same increase in scattering intensity is observed, implying that Li filaments penetrate LLZNO SEs. The fitting parameters show that T_80℃ is significantly larger than T_110℃ and Power-law_80℃ is significantly smaller than Power-law_110℃, implying the thicker and rougher Li filaments. Moreover, no fluctuations in scattering intensity and fitting parameters due to the competition between growth and self-healing were observed. This is related to the weaker self-healing ability of Li at low temperatures. Further explanation will be given in the subsequent theoretical calculations. Besides, to validate the universality of the aforementioned competition between Li filament growth and self-healing in SEs, another typical 77.5Li₂S-22.5P₂S₅ (LPS) electrolyte is also employed using identical operando SANS characterization (Supplementary Fig. 7 and Supplementary Table 5). The XRD pattern of the prepared LPS is shown in Supplementary Fig. 8. As shown in Supplementary Fig. 7a,b, the voltage fluctuates similarly with increasing time, and the synchronous scattering intensity also fluctuates. Furthermore, parameters that reflect microscopic information about the grain boundaries show similar trends (Supplementary Fig. 7c,d). Therefore, it can be concluded that the competition between Li filament growth and self-healing is a universal phenomenon in SEs.

Li filament distribution inside SEs. The SEM was employed to confirm the distribution of Li filaments within the SEs after the short circuit. As shown in Fig. 4a, the LLZNO SEs are composed of densely stacked particles with a diameter of about 1 µm, which are covered by gauze-like flakes. To further verify the gauze-like flakes, a secondary electron image and backscattered electron (BSE) image for the identical area are compared in Fig. 4b. The legible gauze-like flakes in the secondary electron image are almost invisible in the BSE image. This can be rationally explained that the BSE image provides not only morphology but also element distribution, but is insensitive to light elements. It can be concluded that gauze-like flakes consist of light elements. The distribution of O elements is used to trace the distribution of Li filaments since the metal Li is sensitive to O₂ and H₂O.^{12, 39} As shown in Fig. 4c,d, two points for pure LLZNO and gauze-like flakes are compared by EDS analysis. The oxygen content at point 1 is 70.6%, which is consistent with the oxygen content of LLZNO. For gauze-like flakes, the oxygen content increases from 75.1–82.1%, while the content of other elements remains almost constant, confirming that gauze-like flakes are Li filaments. Further EDS mapping shows that O elements are concentrated at the edges of the particles, while Zr and La elements are uniformly distributed, indicating that Li filaments are deposited at the grain boundaries (Fig. 4e-h). In contrast, no gauze-like Li filaments are observed in SEs after only 2 hours of discharge at 5 µA cm^-2 (Supplementary Fig. 9). However, the apparent aggregation of O elements implies that precipitation of Li filaments has occurred in SEs. The resolution of SEM is not high enough to observe them, but operando SANS can capture the nanoscale Li filament variation (Fig. 3). The above postmortem results confirm the existence of Li filaments at the grain boundaries, which is consistent with the SANS results. However, these postmortem results only show the final state of Li filaments without dynamic evolution, not to mention the competition between Li filament formation and self-healing.

Dynamic evolution of Li filaments in SEs and ‘heat therapy’. Using operando SANS with high spatial resolution, the competition between filament formation and self-healing is confirmed. The speculation that enhancing the self-healing effect could further prolong the cell lifespan is thus proposed. Based on recent results, the self-healing effect is supposed to be highly related to temperature.^{34, 36} Additionally, changing working temperature is facile and effective of all the approaches. Thus, the finite element method (FEM) was employed to investigate the temperature effect during the competition between filament formation and self-healing, and to reveal the self-healing mechanism. Firstly, some filaments with acicular shapes are pre-buried in SEs, as exhibited in Fig. 5a. The direction of Li⁺ diffusion is set from the right to the left side. For the scale bar, Ф refers to the volume fraction of Li metal, where 1 represents the pure Li metal phase, while 0 represents the pure SEs without Li metal. With the continuous migration of Li ions, Fig. 5b(ⅰ) shows that Li⁺ is preferentially deposited at the tips of the pre-buried filaments, due to the enhanced electric fields and higher local current densities.^{11, 29} Without intervention, this uneven deposition could accelerate the growth of Li filaments, thus further promoting the dendrite formation and eventually leading to a short circuit. However, at the elevated temperature of 450 K, a counteraction to eliminate the acicular filaments is observed, as pointed out by the black arrows, which is attributed to the high surface energy that makes filament tips shrink and smooth. With further time relaxation, as shown in Fig. 5b(ⅱ) to (ⅱi), although the amount of Li metal still increases, the sharp tips eventually become smooth. This simulated dynamic competition between formation and self-healing at the tips of the filaments is consistent with the above operando SANS observation. By contrast, the dynamic evolution of Li filaments at the room temperature of 300 K is less significant. As shown in Fig. 5c, at the same relaxation time, the formation of Li filaments is observed, while the self-healing is much slower and weaker at the lower temperature. This explains why thicker and rougher Li filaments were detected at 80°C, while no dynamic evolution was observed due to the competition between growth and self-healing (Supplementary Fig. 6 and Supplementary Table 4).

From the simulation results, it can be concluded that the self-healing ability of Li is improved at elevated temperatures, making the growth of Li filaments less "sharp" and effectively slowing down the occurrence of short circuits. Therefore, it is reasonable to conjecture that the heat treatment could positively prevent Li filaments formation. To demonstrate this conception, LiBH₄ SE with a highly reductive nature has been selected for its compatibility and thermodynamical stability toward Li metal. It is worthy to notice here that the Li/LPS and Li/LLZNO interfaces are both unstable or uneven, especially under high temperatures, which are unsuitable for this confirmatory experiment. To verify the stability between Li and LiBH₄, and further determine the “heat therapy” temperature, DSC-TG analysis has been employed, as illustrated in Supplementary Fig. 10. There is only one obvious peak corresponding to the orthorhombic–hexagonal phase transformation of LiBH₄ from the orthorhombic phase below 170℃, with no weight change throughout. Only the hexagonal P6₃mc phase provides high conductivity up to 10^− 3 S cm^− 1.⁴⁰ The critical current density (CCD) of SEs is an important parameter to evaluate the ability to suppress the Li dendrites. At the lower temperature of 125°C, the voltage first increases as the current density increases, and then suddenly drops at 2.6 mA cm^− 2 (Supplementary Fig. 11). Hence, the CCD value at 125°C is determined as 2.6 mA cm^− 2. Increasing the temperature from 125°C to 140°C and 160°C can effectively promote the CCD values of Li|LiBH4|Li cells to 3.6 and 4.0 mA cm^− 2, respectively (Supplementary Figs. 12,13). A similar observation has been recently reported by Hu et al,¹⁵ who indicate that the higher conductivity at a high temperature increases the amount of Li⁺ near the tip of Li dendrites, thus leading to high surface energy, making filament tips shrink and smooth.

Furthermore, to verify the feasibility of heat therapy in prolonging the cyclability, some Li filaments were pre-buried in Li|LiBH₄|Li cell by initially cycling for 5 cycles at 2 mA cm^− 2, as shown in Supplementary Fig. 14. Subsequently, this cell was subjected to heat therapy under 160°C for 6 hours. then cooled to 125°C for 3 h and cycled again at 2 mA cm^− 2. The symmetric cell can run more than 130 cycles after heat therapy. In sharp contrast, another Li|LiBH₄|Li cell without high-temperature heat therapy can only operate 38 cycles under the same condition (Supplementary Fig. 15). Herein, these results not only strongly confirmed the role of high temperature played in preventing Li filament formation, but also proposed a brand-new and viable heat therapy method to prolong the lifespan of ASSLMBs.

In conclusion, for the first time, we achieved real-time detection of the initial growth of Li filaments at the nanoscale using operando SANS, which provides another perspective for understanding the Li filament formation mechanism. The previous opinion insists that Li filament formation is an irreversible process just like end-stage cancer in ASSLMBs. Different from this opinion, with the aid of operando SANS and theoretical simulations, we have noticed that instead of the single accumulation of Li filaments, there exists competition between Li filament growth and self-healing. This dynamic competition is highly dependent on temperature changes, and the enhanced self-healing ability at elevated temperatures plays a positive role in preventing Li filament formation and can even eliminate already existing Li filaments. This means that the malignant Li filament formation can be possibly ameliorated or even partially eliminated, so long as the Li filaments can be detected early enough. To meet the demand for early detection, operando SANS with high spatial resolution is a powerful tool, providing a new platform for studying Li filaments in SEs. Based on the discovery of dynamic competition, a new method of heat therapy will be a promising route to increase the cycle life of solid-state batteries.

Preparation of Li₆.₅Laz₁.₅Nb₀.₅O₁₂ (LLZNO). The overall preparation of LLZNO is divided into four steps: mixing, pre-sintering, pressing, and sintering. Firstly, stoichiometric amounts of Li₂CO₃ (more than 15% to replenish lithium volatilization during sintering), La₂O₃, ZrO₂, and Nb₂O₅ were added to a ceramic ball mill tank and ball milled for 24 hours at 100 r.p.m. using anhydrous ethanol as the solvent. The mixed slurry was then dried, sieved, and placed in a MgO ceramic crucible for pre-sintering in a muffle furnace at 950°C for 6 h in air. The pre-sintering powder was then removed and ball milled, consistent with the previous ball milling method. Then, the powder was pressed at a pressure of about 6 MPa into pellets with a diameter of about 20 mm. Finally, the pellets buried in the powder were placed in a MgO ceramic crucible and sintered in a muffle furnace at 1250°C for 40 min.

Preparation of 77.5Li₂S-22.5P₂S₅ (LPS). The LPS was prepared by high-energy ball milling. In the glove box, 178.25 mg of Li₂S and 249.75 mg of P₂S₅ were placed in an agate tank, and agate beads were added at a ball-to-powder mass ratio of 60:1. Seal the tank in the glove box and ball mill at 650 r.p.m. The powder is removed from the tank every 5h and ground in a glove box, followed by ball milling. Repeated this operation twice, and continued to ball mill for 24h. Finally, the material was taken out in the glove box, ground, and cold-pressed into pellets with a diameter of about 20 mm.

Characterization. The Bruker D8 Advance X-ray diffractometer with Cu Kα radiation (λ = 1.5406 Å) was used for XRD testing. For air-sensitive LPS SEs, sealed with clear tape before testing and transferred quickly. The XRD pattern was refined by program RIETAN-2000 of the Rietveld method. The morphologies of the samples were investigated by a field-emission SEM (FEI Helios G4 CX). For the discrimination of heavy and light elements in the samples, the backscattered electron (BSE) function of SEM was used. The element distribution was drawn by dual Bruker 30 mm^− 2 detectors’ EDS mapping.

Electrochemical Measurement. The assembly of Cu|LLZNO|Li and Cu|LPS|Li relies on the self-developed device. Specifically, the Cu ring, SEs, and Li foil were stacked one on top of the other and then connected to the external circuit by wires. Before the electrochemical test, the cells were held at the specified test temperature for 1 h to ensure uniform heat distribution throughout the cells. The Cu|SEs|Li cells were kept in the sealed device filled with argon gas during the whole testing. For Cu|LLZNO|Li, the temperature was controlled at 110 ± 8°C to ensure high ionic conductivity, while for Cu|LPS|Li, the temperature was controlled at 80 ± 5°C (Supplementary Fig. 16). The open-circuit voltage curves of Cu|LLZNO|Li and Cu|LPS|Li are shown in Supplementary Fig. 17. Li|LPS|Li and Li|LiBH₄|Li were assembled by attaching lithium foils to both sides of the pressed SEs pellets. At a pressure of 360 MPa, 150 mg of prepared LPS and 120 mg of purchased LiBH₄ were pressed into pellets, respectively. Assembly of all cells is done in the glove box, eliminating exposure to air. Electrochemical tests were performed on an electrochemical workstation (CHI 760e) and a battery test system (BTS 7.6.0.1313, NEWARE). The electrochemical impedance spectrum (EIS) measurement was performed on an electrochemical workstation (CHI 760e). The EIS was measured in the frequency range from 0.1 to 10⁶ Hz.

SANS measurement. Operando SANS measurements were performed at the China Mianyang Research Reactor (CMRR), on the SANS-Suanni spectrometer. The sample-to-detector distance (SDD) is 1.8 m and the wavelength λ is 0.53 nm, resulting in a q range of 0.2 to 2.0 nm^− 1.⁴¹ The magnitude of the scattering vector q defined by $q=(4\pi /\lambda )\text{s}\text{i}\text{n}(\theta /2)$, where θ is the scattering angle. The wavelength spread (Δλ/λ) is currently about 18%. After installing the cell in the device, the device was installed at the sample desk. After holding at a specific temperature for one hour, signal acquisition began. The scattering signal is obtained by subtracting the background scattering signal from the total scattering signal, which is mainly done by the BerSANS software. The specific data fitting analysis was performed using Igor Pro software.⁴²

Theoretical calculations. The finite element method (FEM) is applied through Comsol Multiphysics, version 5.6. For the self-healing mechanism, the 2D level set equation was used to describe the evolution of dendrite morphology:

$$\frac{\partial \varphi }{\partial t}+\overrightarrow{u}\cdot \nabla \varphi =\gamma \nabla \left[{ϵ}_{ls}\nabla \varphi -\varphi \left(1-\varphi \right)\frac{\nabla \varphi }{\left|\nabla \varphi \right|}\right]$$

where Ф refers to the volume fraction of Li metal, which is equal to 1 in the pure metal phase and between 0 and 1 at the interface. The external phase interface velocity u is powered by the deposition reaction of Li ions, which is not the focus of our simulation thus u was set to be a constant value along the x-direction. The coefficient γ controls the internal evolution velocity u_n = -γk, which is normal to the interface and proportional to the interface curvature k. It describes the self-diffusion behavior of the dendrite interface under a linear approximation which is in favor of a low-curvature morphology. Since our purpose is to simulate the thermal-diffusion behavior which is insignificantly related to the electric driving, we approximately ignored the electrical effect and simplified the dendrite formation process from electrochemical reaction to liquid-solid freezing. Then according to the study from Sun et.al.,⁴³ γ can be written as:

$$\gamma \propto {C}_{P}D$$

where Cp is the heat capacity and D is the thermal conductivity of Li metal. For regions close to room temperature (300 K) and below the Debye temperature of lithium (344 K), the Debye approximation is applicable. Based on this, both Cp and D are proportional to T³ and the coefficient γ is proportional to T⁶. The initial morphology of the Li dendrites was generated by an arbitrary function. The left boundary was set inlet condition while the right boundary was set outlet. The upper and lower boundaries were set to be no-flux. The whole system represented a surface of electrolyte particles with a system size of 50µm. The default mapped mesh was applied with fluid-dynamics type and extremely fine size level. The default phase-initialization study step and time-dependent solver were used. The mesh optimization was studied and our simulation result showed insignificant dependence on meshing.

Acknowledgments

This work is financially supported by the National Natural Science Foundation of China (52071083, 51871059, 51922031, and 52071086), the National Major Scientific Research Equipment of China (No. 51727801), the National Key R&D Program of China (No. 2018YFB1502105), the Science and Technology Commission of Shanghai Municipality (No. 20ZR1405400), and Pujiang Talent Program of Shanghai (20PJ1401400).

Competing interests

The authors declare no conflict of interest.

1. Zhou Q, Ma J, Dong S, Li X, Cui G. Intermolecular Chemistry in Solid Polymer Electrolytes for High-Energy-Density Lithium Batteries. Adv Mater 31, e1902029 (2019).

2. Janek J, Zeier WG. A solid future for battery development. Nat Energy 1, 16141 (2016).

3. Monroe C, Newman J. The effect of interfacial deformation on electrodeposition kinetics. Journal of the Electrochemical Society 151, A880-A886 (2004).

4. Monroe C, Newman J. The impact of elastic deformation on deposition kinetics at lithium/polymer interfaces. Journal of the Electrochemical Society 152, A396-A404 (2005).

5. Cheng XB, Zhang R, Zhao CZ, Zhang Q. Toward Safe Lithium Metal Anode in Rechargeable Batteries: A Review. Chem Rev 117, 10403-10473 (2017).

6. Porz L, et al. Mechanism of Lithium Metal Penetration through Inorganic Solid Electrolytes. Adv Energy Mater 7, 1701003 (2017).

7. Famprikis T, Canepa P, Dawson JA, Islam MS, Masquelier C. Fundamentals of inorganic solid-state electrolytes for batteries. Nat Mater 18, 1278-1291 (2019).

8. Doux JM, et al. Stack Pressure Considerations for Room‐Temperature All‐Solid‐State Lithium Metal Batteries. Adv Energy Mater 10, (2019).

9. Yu ZJ, et al. Dendrites in Solid-State Batteries: Ion Transport Behavior, Advanced Characterization, and Interface Regulation. Adv Energy Mater 11, 2003250 (2021).

10. Liu X, et al. Local electronic structure variation resulting in Li 'filament' formation within solid electrolytes. Nat Mater 20, 1485-1490 (2021).

11. Ning Z, et al. Visualizing plating-induced cracking in lithium-anode solid-electrolyte cells. Nat Mater 20, 1121-1129 (2021).

12. Han FD, et al. High electronic conductivity as the origin of lithium dendrite formation within solid electrolytes. Nat Energy 4, 187-196 (2019).

13. Wang C, et al. In Situ Neutron Depth Profiling of Lithium Metal-Garnet Interfaces for Solid State Batteries. J Am Chem Soc 139, 14257-14264 (2017).

14. Kazyak E, et al. Li Penetration in Ceramic Solid Electrolytes: Operando Microscopy Analysis of Morphology, Propagation, and Reversibility. Matter 2, 1025-1048 (2020).

15. Guo WC, et al. In-situ optical observation of Li growth in garnet-type solid state electrolyte. Energy Storage Mater 41, 791-797 (2021).

16. Westover AS, Dudney NJ, Sacci RL, Kalnaus S. Deposition and Confinement of Li Metal along an Artificial Lipon-Lipon Interface. Acs Energy Letters 4, 651-655 (2019).

17. Chandrashekar S, Trease NM, Chang HJ, Du LS, Grey CP, Jerschow A. 7Li MRI of Li batteries reveals location of microstructural lithium. Nat Mater 11, 311-315 (2012).

18. Bhattacharyya R, Key B, Chen H, Best AS, Hollenkamp AF, Grey CP. In situ NMR observation of the formation of metallic lithium microstructures in lithium batteries. Nat Mater 9, 504-510 (2010).

19. Harry KJ, Hallinan DT, Parkinson DY, MacDowell AA, Balsara NP. Detection of subsurface structures underneath dendrites formed on cycled lithium metal electrodes. Nat Mater 13, 69-73 (2014).

20. Tan DHS, Banerjee A, Chen Z, Meng YS. From nanoscale interface characterization to sustainable energy storage using all-solid-state batteries. Nat Nanotechnol 15, 170-180 (2020).

21. Li Q, et al. In-situ visualization of lithium plating in all-solid-state lithium-metal battery. Nano Energy 63, 103895 (2019).

22. Krauskopf T, et al. Lithium-Metal Growth Kinetics on LLZO Garnet-Type Solid Electrolytes. Joule 3, 2030-2049 (2019).

23. Gong Y, et al. In Situ Atomic-Scale Observation of Electrochemical Delithiation Induced Structure Evolution of LiCoO2 Cathode in a Working All-Solid-State Battery. J Am Chem Soc 139, 4274-4277 (2017).

24. Zhu F, et al. Single-atom-layer traps in a solid electrolyte for lithium batteries. Nat Commun 11, 1828 (2020).

25. Jung H, et al. Nanoscale in situ detection of nucleation and growth of Li electrodeposition at various current densities. J Mater Chem A 6, 4629-4635 (2018).

26. Li J, et al. Neutron diffractometer RSND for residual stress analysis at CAEP. Nuclear Instruments & Methods in Physics Research Section a-Accelerators Spectrometers Detectors and Associated Equipment 783, 76-79 (2015).

27. Hofmann M, et al. The new materials science diffractometer STRESS-SPEC at FRM-II. Physica B: Condensed Matter 385-386, 1035-1037 (2006).

28. Awaka J, Takashima A, Kataoka K, Kijima N, Idemoto Y, Akimoto J. ChemInform Abstract: Crystal Structure of Fast Lithium-Ion-Conducting Cubic Li7La3Zr2O12. ChemInform 42, (2011).

29. Swamy T, et al. Lithium Metal Penetration Induced by Electrodeposition through Solid Electrolytes: Example in Single-Crystal Li6La3ZrTaO12 Garnet. Journal of the Electrochemical Society 165, A3648-A3655 (2018).

30. Hammouda B. Probing Nanoscale Structures—The SANS Toolbox. National Institute of Standards and Technology Center for Neutron Research (2016).

31. Hammouda B. A new Guinier-Porod model. Journal of Applied Crystallography 43, 716-719 (2010).

32. Schmidt PW. Small-Angle Scattering Studies of Disordered, Porous and Fractal Systems. Journal of Applied Crystallography 24, 414-435 (1991).

33. Kohls DJ, Beaucage G. Rational design of reinforced rubber. Curr Opin Solid State Mat Sci 6, 183-194 (2002).

34. Li YQ, et al. Self-Healing Properties of Alkali Metals under "High-Energy Conditions" in Batteries. Adv Energy Mater 11, 2100470 (2021).

35. Parejiya A, Amin R, Essehli R, Wood DL, Belharouak I. Electrochemical Healing of Dendrites in Garnet-Based Solid Electrolytes. Acs Energy Letters 5, 3368-3373 (2020).

36. Li L, et al. Self-heating-induced healing of lithium dendrites. Science 359, 1513-1516 (2018).

37. Hundekar P, et al. Exploiting self-heat in a lithium metal battery for dendrite healing. Energy Storage Mater 20, 291-298 (2019).

38. Hundekar P, et al. In situ healing of dendrites in a potassium metal battery. Proceedings of the National Academy of Sciences 117, 5588-5594 (2020).

39. Cheng EJ, Sharafi A, Sakamoto J. Intergranular Li metal propagation through polycrystalline Li6.25Al0.25La3Zr2O12 ceramic electrolyte. Electrochim Acta 223, 85-91 (2017).

40. Mo F, et al. Inside or Outside: Origin of Lithium Dendrite Formation of All Solid‐State Electrolytes. Adv Energy Mater 9, (2019).

41. Tian Q, et al. Calibration of the Suanni small-angle neutron scattering instrument at the China Mianyang Research Reactor. J Appl Crystallogr 51, 1662–1670 (2018).

42. Kline SR. Reduction and analysis of SANS and USANS data using igor. J Appl Crystallogr 39, (2006).

43. Sun Y, Beckermann C. Sharp interface tracking using the phase-field equation. Journal of Computational Physics 220, 626-653 (2007).

There is NO Competing Interest.

SupplementaryInformation.docx
Revealing the dynamic evolution of Li filaments within solid electrolytes by operando small-angle neutron scattering

Download PDF

Version 1

posted

You are reading this latest preprint version

Revealing the dynamic evolution of Li filaments within solid electrolytes by operando small-angle neutron scattering

Status:

Version 1

Abstract

Figures

Introduction

Results And Discussion

Methods

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1