Garnet-based solid state batteries benefitting from an ionic/electronic mixed conductive interface constructed by lithiation of porous FeS₂

The rapid development of portable electronic devices, electric vehicles, and distributed energy storage grids places an increasingly urgent demand on lithium-based batteries with high energy density, long life, and high safety.¹ However, the traditional lithium-ion batteries based on a graphite anode and an organic electrolyte cannot meet the need of high energy density and also face the risk of electrolyte flammability and leakage.² High-performance solid-state batteries (SSBs) are considered as the candidates for the next generation of lithium batteries because they can be matched with Li metal anodes and high-voltage cathodes to achieve the dual upgrades of energy density and safety.³ As a key component, a solid electrolyte plays a crucial role in the electrochemical performance of SSBs. Currently, various inorganic solid electrolytes have been developed, such as sulfide glass ceramics, perovskite oxides, NASICON-type oxides (NASICON is a Na⁺ superionic conductor), and garnet-type oxides.^4,5 Among these inorganic solid electrolytes, the garnet-type solid electrolyte Li₇La₃Zr₂O₁₂ (LLZO) has the high ionic conductivity (10⁻³ S cm⁻¹) and good chemical and electrochemical stabilities to lithium metal anode⁶ and has been extensively studied.

However, the surface of garnet-type LLZO ceramic electrolytes is lithiophobic and has the poor wettability to lithium metals, leading to the formation of a point-to-point contact between LLZO and Li. The heterogeneity of interface contact increases the Li/LLZO interface resistance and can also lead to the excessive local current density,⁷ which further causes the excessive concentration of Li⁺ flux and uneven electric field distribution. Once the local potential reaches the deposition potential of lithium, Li ions acquire electrons to form the initial deposits of lithium, which are often the nucleation sites for lithium dendrites. Subsequently, lithium dendrites further spread and penetrate the grain boundaries of LLZO, triggering the continued nucleation of lithium and additional electron injection.⁸ Therefore, improving the interface contact performance between Li and LLZO plays a crucial role in guiding the uniform lithium deposition and inhibiting the formation of lithium dendrites.

In this regard, various interfacial layers have been developed to solve the interface problems of garnet electrolytes. Previously developed interface modifications were mainly divided into the improvements of electronic conductivity and ion conductivity. For the electron-conducting intermediate layer, various lithium alloyable interfaces have been developed that can effectively improve the microstructural properties of the Li/garnet interface and enhance its wettability. For example, building a Li–Sn alloy buffer layer at the Li/garnet interface can effectively improve the interface wettability and inhibit lithium dendrites through the self-limiting reaction between the Sn thin layer and the Li metal.⁹ Other alloyable interlayers, such as Na,¹⁰ Ge,¹¹ and Cu₅Sn₆,¹² have also been reported. Subsequent studies found that the potential electronic conductivity in LLZO is the origin of the penetration of lithium dendrites into the grain boundaries of solid electrolytes.¹³ Especially under high current densities, lithium ions can preferentially accumulate at the grain boundaries and defects of LLZO ceramics, leading to the formation of lithium dendrites and battery performance degradation. Therefore, in order to avoid electrons jumping into the garnet solid electrolyte at the anode interface, introducing an ion-conducting Li solid electrolyte into the interface layer seems to be a good choice. This type of interface layer has negligible electronic conductivity and prevents electrons from attacking the electrolyte during lithium deposition. For example, Li₃N prepared by plasma-enhanced chemical vapor deposition was used to coat the surface of a garnet solid electrolyte. Due to the high Li-ion conductivity of Li₃N, the corresponding solid state Li symmetric cells can be cycled stably.¹⁴ Although the single Li-ion conductor interface layer provides a new solution for Li/garnet ceramic interface modification, it may cause a large interface resistance between the Li and solid electrolyte interface, resulting in an increased overpotential in symmetrical cell cycling and eventually battery failure.¹⁵ Therefore, relying on a purely ion-conducting or purely electronic-conducting interlayer is not sufficient to maintain the long-term stability of interface. With this in mind, building the hybrid ion/electron conductive layers seems to be a more promising option. Currently, a variety of hybrid conductors have been used for the interface layer between lithium and garnet ceramics. For example, a Cu₃N thin layer was constructed on the surface of a solid electrolyte through magnetic sputtering.¹⁶ This interface layer enables the formation of an ion/electron conductive Li₃N/Cu composite layer after lithiation. However, this expensive fabrication process seems to be difficult to perform adequately in practical applications. Therefore, fast, convenient, and effective interface modification strategies are needed to meet the practical application of garnet-based electrolytes. Furthermore, the dense thin layers (such as those obtained by physical deposition methods) easily cause the local mechanical expansion during lithiation process and weaken the interface wettability effect of the interface layer. In this regard, the porous interface layer is more conducive to the melt penetration of lithium and ensures the compactness of the interface during subsequent cycles.

Here, in order to enhance the contact between the Li metal and Ta doped LLZO (denoted as LLZTO, i.e., Li_6.5La₃Zr_1.5Ta_0.5O₁₂) interface and suppress Li dendrites, we successfully constructed a porous FeS₂ coating layer. The advantage of using the electrode material as the modification interface is that the structure is relatively stable, and it can ensure the efficient transport of lithium ions within its bulk phase. Compared with elemental sulfur or other sulfide materials, FeS₂ has a much higher intrinsic electronic conductivity, which can reach more than 10⁻⁶ S cm⁻¹ at room temperature, and can establish an internal conductive network during the electrochemical reaction process. FeS₂ synthesized by a solvothermal method has a porous network structure. During the conversion reaction between FeS₂ and Li, more Li is filled inside the pores. The potential formation of lithiophilic Fe nanodomains helps to confine Li inside the pores, thereby leading to the formation of a good Li/FeS₂-LLZTO interface contact with high stability. In addition, the formed Li₂S matrix with a defect structure, as a potential Li-ion conductive network, can mitigate the electronic tuning effect and prevent the growth of lithium dendrites at the Li/LLZTO interface and inside LLZTO. Due to the synergistic effect of Li₂S and Fe mixed conductive nanodomains, the lithium plating/stripping performance exhibits the long-term durability in both lithium symmetric cells and LiFePO₄/Li full cells.

As shown in the x-ray diffraction (XRD) pattern in Fig. 1(a), the FeS₂ sample synthesized by using the solvothermal method is highly amorphized and does not exhibit distinct diffraction peaks. Only three low-intensity peaks appear at 18.67°, 27.86°, and 49.08°, indicating a low crystallinity of an as-synthesized sample. This defective structure is favorable for the formation of porous texture and the infiltration of melted Li metal, promoting its conversion reaction with the formation of rich Li₂S/Fe nanodomain interfaces. As shown in Fig. 1(b), the microstructure of synthesized FeS₂ is observed by scanning electron microscopy (SEM) imaging. It reveals that FeS₂ has a porous structure with abundant macropores, which can accommodate a significant amount of Li when in contact with molten Li. Therefore, coating FeS₂ on the surface of LLZTO can enhance the wettability of Li on the LLZTO surface. X-ray photoelectron spectra (XPS) were also employed to investigate the chemical states of related elements of a sample surface, and the spectra of Fe 2p and S 2p are shown in Figs. 1(c) and 1(d). In Fig. 1(c), the peaks at the binding energies of 708.03 and 721.5 eV correspond to the splitting peaks of Fe 2p_3/2 and Fe 2p_1/2 of Fe²⁺ in FeS₂. The peak at 710.58 eV is attributed to the Fe³⁺ species, which can be ascribed to the surface oxidation of FeS₂ when exposed to air. The satellite peaks at 714.03 and 730.53 eV correspond to the hybridization of Fe²⁺ species in Fe 2p_3/2 and Fe 2p_1/2, respectively.^17,18 In Fig. 1(d), the paired split peaks at binding energies of 161.78 eV (S 2p_3/2) and 163.83 eV (S 2p_1/2) coincide with the peaks of the S₂²⁻ dimer coordinated with Fe²⁺, while the weaker peaks at 167.48 and 169.18 eV correspond to the signals of SO₄²⁻, also indicating the surface oxidation of FeS₂.^17,18 A small amount of Fe³⁺ and SO₄²⁻ is located on the shallow surface of FeS₂, and it would not affect the porous structure and lithiophilicity of FeS₂ skeleton as discussed later.

Due to the porous structure of FeS₂, the powder sample is relatively fluffy and it can be evenly painted on the LLZTO surface with a doctor blade. The excess and poorly adhered powder can be blown off from the LLZTO surface. As shown in Fig. 2(a), the LLZTO surface becomes dark yellow after the coating of amorphized FeS₂. The coated LLZTO is then immersed in molten Li, and the changes on the electrolyte pellet surface are observed [Figs. 2(b)–2(f)]. The lithiation state of FeS₂ can be judged by the color change. After the initial contact, a small amount of Li is infiltrated to the FeS₂-LLZTO surface. As the number of this operation increases, the area covered by Li gradually expands. Due to the presence of numerous pores in FeS₂, the pores are prone to be filled with Li upon infiltration. The process of Li infiltration into FeS₂-LLZTO is also accompanied by the gradual lithiation of FeS₂ based on the conversion reaction with the formation of Li₂S and Fe [Fig. 2(g)]. The in situ generated Li₂S is expected to be also amorphized and defective, and therefore, it can serve as a lithium-ion conductor with an ion conductivity of up to 10⁻⁵ S cm⁻¹.^19,20 The in situ generated Fe domains are expected to be lithiophilic and can disperse the Li nucleation sites and plating mass. The synergistic effect of an in-situ-formed Li₂S and Fe mixed-conducting layer enables the homogenization of Li-ion flowing and Li mass plating. During the subsequent Li deposition/stripping process, the porous structure and buffering effect of Li₂S would mitigate the volume expansion and allow the interface to maintain the tight contact.

Since it is difficult to completely remove Li from the electrolyte surface after reacting with FeS₂-LLZTO, we used the FeS₂ powder to verify the composition of conversion products after lithiation reaction. We stuffed the synthesized FeS₂ powder with molten Li and kept the lithiation process for 30 min to simulate the interface reaction process, and then, the composition of the resulting products was measured by using XPS. To minimize the interference from the oxidation of Fe and S elements in air before testing, the samples were etched for 60 s prior to testing. As shown in Figs. 2(h) and 2(i), the Fe 2p spectrum has two split peaks at 707.00 and 718.85 eV, corresponding to the signals of metallic Fe⁰. There are no other peaks presented besides Fe⁰, indicating that the Fe²⁺ species in the sample after lithiation has been completely reduced to metallic Fe.²¹ The S 2p spectrum also has two split peaks at 161.00 and 162.25 eV, corresponding to the S signal in Li₂S.²² The XPS test results confirm that the molten Li reacts with FeS₂-LLZTO to form a Li₂S/Fe interface layer.

The interface morphology of FeS₂-LLZTO and Li/FeS₂-LLZTO can be clearly observed under SEM imaging, as shown in Fig. 3(a). It can be seen that FeS₂ can form a good contact with LLZTO, which is a prerequisite for the close contact between Li and FeS₂-LLZTO. Part of FeS₂ is also filled in the depressions on the LLZTO surface. This portion of FeS₂ contributes to the penetration of Li into the LLZTO surface, avoiding the interruption of Li⁺ transmission channels caused by interfacial gaps during the electrochemical cycle. As shown in Fig. 3(b), the thickness of the FeS₂ coating layer is ∼1.43 μm, and the nanostructure of FeS₂ is favorable for its thinning and conformal coating effects. The thickness of the modified layer can be modulated by the amount of FeS₂ powder. Generally, the interface layer is optimally kept at 1–2 μm. From the energy dispersive spectra (EDS) mapping analysis of element distribution at the cross section of FeS₂-LLZTO [Fig. 3(c)], Fe and S are mainly distributed in the FeS₂ layer. It is noteworthy that the nanostructure and porous morphology of FeS₂ are still observed on the LLZTO interface after coating, favorable for the following sufficient lithiation and Li mass stuffing. The cross-sectional morphology of Li/FeS₂-LLZTO reveals a well-formed interface contact between Li and LLZTO without apparent gap [Fig. 3(d)]. There is an intermediate layer with a distinct morphology from Li to LLZTO. The thickness of this intermediate layer is measured to be ∼1.461 μm, which is very close to the thickness of the pristine FeS₂ layer on the LLZTO surface. This further indicates that sufficient Li penetrates into the pores of nanostructure FeS₂. It is important to mention that the main components of the intermediate transition layer at this time are Li₂S and Fe [Figs. 3(e)–3(j)]. This ion/electron mixed conductive layer can enhance the Li ion flux in the transition region, ensuring the smooth transport of Li⁺ at the interface. It would effectively reduce the interface resistance at the anode/electrolyte interface and suppress the growth of lithium dendrites.

The resistance evolution of the Li/FeS₂-LLZTO/Li cell at room temperature and 60 °C was tested by electrochemical impedance spectra (EIS), and the interface resistance is calculated, as shown in Fig. 4(a). The impedance spectra for the Li/FeS₂-LLZTO interface contain two semicircles. The calculated area-specific resistance (ASR) for the single-sided Li/FeS₂-LLZTO interface is 26.83 Ω cm² at room temperature and 15.20 Ω cm² at 60 °C, significantly lower than that of the unmodified Li/LLZTO interface (about 310.9 Ω cm²). This comparison effect is mainly attributed to the excellent interface contact in Li/FeS₂-LLZTO and the formation of the mixed-conducting transition layer after the lithiation of FeS₂. It is worth noting that the modification effect of this method is superior to the reported similar modification of the Li/LLZO interface using MoS₂ as an intermediate transition layer (with the ASR of 14 Ω cm² at 100 °C).²³ The symmetric Li/FeS₂-LLZTO/Li cell exhibits the excellent rate capability [Fig. 4(b)]. The polarization voltages at the current densities of 200, 300, 400, and 600 μA cm⁻² are 9.0, 15.5, 21.7, and 37.8 mV, respectively. When the current density is reduced to 200 μA cm⁻², the polarization voltage also returns to 9.0 mV. Long-term cycling tests of the symmetric cells at 100 and 400 μA cm⁻² are shown in Figs. 4(c) and 4(d). At 100 μA·cm⁻², the stable cycling is maintained for over 1150 h, with the polarization voltages of 12.1, 17.4, and 16.7 mV in the fifth, 500th, and 600th cycles, respectively. Stable cycling is also achieved for over 400 h at 400 μA cm⁻². At room temperature, the symmetric cell exhibits a polarization of 13.6 mV at a current density of 50 μA cm⁻² (Fig. S1) and an initial polarization voltage of 30.1 mV at 100 μA cm⁻², with a stable cycling for over 450 h. Thus, it is evident that the Li/FeS₂-LLZTO interface is highly stable and effectively inhibits the growth of lithium dendrites. In contrast, the Li/LLZTO/Li control cell shows much larger overpotential during the initial cycling and reaches short circuit soon under 100 μA cm⁻² (Fig. S2). These phenomena are associated with the uneven Li plating and stripping caused by the poor contact of Li with LLZTO.

The ionic conductivity and electronic conductivity are crucial to the performance of the FeS₂ intermediate layer. Due to the direct contact between Li and FeS₂ at the Li/FeS₂-LLZTO interface, it is challenging to directly test the conductivity of lithiated FeS₂. We assembled a Li–FeS₂ equivalent cell to simulate the electrochemical behavior at the interface, and the electrolyte used is the commercial LiPF₆ electrolyte. The galvanostatic intermittent titration technique (GITT) was employed to investigate the Li⁺ diffusion kinetics during Li⁺ insertion into FeS₂, thereby estimating the ionic conductivity of the FeS₂ intermediate layer.^24,25 The result is shown in Fig. 5(a). Based on the GITT curves, the quantitative relationship between the Li⁺ diffusion coefficient (D_Li+) and voltage at 60 °C can be calculated. When the discharge voltage is 1.2 V, FeS₂ is fully lithiated and the calculated D_Li+ at this point is ∼2.51 × 10⁻¹² cm² s⁻¹ [Fig. 5(b)]. According to the Nernst–Einstein equation and D_Li+,^23,26 the calculated ionic conductivity of fully lithiated FeS₂ is ∼1.58 × 10⁻⁶ S cm⁻¹ at room temperature. This indicates that the ionic conductivity of the transition layer is high enough, facilitating the uniform transport and Li⁺ and deposition/stripping of Li at the interface.

We also assembled the Li/FeS₂-LLZTO/LiFePO₄ solid-state cells and tested their cycling and rate performance, and the results are shown in Figs. 5(c) and 5(d). After activating and stabilizing the internal interfaces by five cycles at a current density of 50 μA cm⁻², the first-cycle capacity of the full cell reaches 168.1 mAh g⁻¹ at a current density of 100 μA cm⁻², and the stable cycling is maintained for 50 cycles. The capacity in the 50th cycle remains at 125.9 mAh g⁻¹, and the Coulombic efficiency can be maintained at around 99%. In the rate performance test, the cell exhibits the capacities of 165.4, 158.1, 147.1, 132.6, 128.3, and 123.4 mAh g⁻¹ at the current densities of 150, 200, 250, 300, 350, and 400 μA cm⁻², respectively. When the current density is reduced to 100 μA cm⁻², the cell capacity recovers to 166.7 mAh g⁻¹, demonstrating the excellent electrochemical reversibility and high-rate tolerance. The outstanding performance of full cells indicates the good contact at various interfaces within the solid state battery architecture.²⁷ The stable high-flux ion transfer at the Li/FeS₂-LLZTO interface enables the solid-state battery to exhibit the excellent reversible capacity even at a large current density of 400 μA cm⁻².²⁸ 

In this work, a porous amorphized FeS₂ was developed as a Li/LLZTO modified layer through a hydrothermal method. The contact between the Li anode and LLZO interface can be reinforced through the tight and continuous Li₂S/Fe mixed ion/electron conductive layer through the conversion reaction between porous FeS₂ and molten lithium. The lithiophilic Fe domains distribute electrons and lithium nucleation sites evenly at the garnet/Li interface, while defective Li₂S acts as a fast ion conductor to ensure the rapid and effective transport of Li⁺ at the garnet/Li interface. The synergistic effect of Fe and Li₂S enables the uniform Li deposition at the garnet/Li interface, thereby inhibiting the growth of lithium dendrites. The resistance of the Li/FeS₂-LLZTO interface at 60 °C is as small as 15.20 Ω·cm², and the ion conductivity of fully lithiated FeS₂ is estimated to be 1.58 × 10⁻⁶ S·cm⁻¹ at room temperature. The converted intermediate transition layer has good dynamic stability and can effectively reduce the voltage polarization during battery cycling. The Li/FeS₂-LLZTO/Li symmetric cell can cycle stably for more than 1150 h at a current density of 100 μA·cm⁻² with a polarization voltage not exceeding 20 mV and can also cycle stably for more than 400 h at a much higher current density of 400 μA·cm⁻². This simple and feasible interface modification strategy and its excellent lithium dendrite suppression effect have good application prospects in the design of garnet-based solid-state lithium metal batteries.

See the supplementary material for the experimental section, Figs. S1 and S2.

This work was supported by the National Natural Science Foundation of China (Grant Nos. 52372249, 21975276, and 52102329). C. Li appreciates the support from the Program of Shanghai Academic Research Leader (Grant No. 21XD1424400).

The authors have no conflicts to disclose.

Y.Z. and Y.Z. contributed equally to this work.

Yuhan Zeng: Data curation (equal); Formal analysis (equal); Investigation (equal). Yang Zhang: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal). Jiulin Hu: Methodology (equal); Supervision (equal); Writing – original draft (equal). Chilin Li: Conceptualization (equal); Funding acquisition (equal); Methodology (equal); Resources (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article and its supplementary material.