Chemical design rules for low-resistivity electrode–electrolyte interfaces in all-solid-state lithium batteries

Abstract

A key issue of all-solid-state rechargeable lithium batteries is the formation of a low-resistive electrode‒solid electrolyte interface. Since this problem begins with undesirable side reactions during the electrode‒solid electrolyte interface bonding process, the basic guidelines for suppressing side reactions need to be clarified. Here, we combine a typical Li⁺-conductive amorphous solid electrolyte, lithium phosphate, which exhibits various Li chemical potentials depending on the Li/P atomic ratio on the film surface, with c-axis oriented LiCoO₂ thin films. The lowest interfacial resistances of less than 10 Ω cm² are achieved within a certain range of Li/P atomic ratios. We clarify that interface bonding is classified as Li-insertion or Li-extraction types on the electrodes. The formation of appropriate ranges of Li-inserted or Li-extracted electrode states at the bonding interface is a major factor for attaining low-resistivity electrode–solid electrolyte interfaces and determining their appropriate combinations.

Introduction

All-solid-state lithium batteries (SSBs) are promising power sources for advanced electric vehicles, satellites, and surface-mounted semiconductor devices because of their high-power density, liquid-free system, and enhanced safety and thermal stability features^1,2. Two critical issues in enhancing the performance of SSBs are as follows: the solid electrolyte (SE) and the interface. The former issue is related to the development of solid electrolytes (SEs) with high lithium-ion (Li⁺) conductivity. Research aimed at discovering highly Li⁺ conductive SEs has been conducted based on experimental and theoretical approaches^3,4,5,6. Researchers have discovered sulfide³, oxyhalide⁴, and composite⁵ SEs that achieve Li⁺ conductivities of approximately 10 mS cm⁻¹ or higher at room temperature and have proposed design principles for these superionic conductors⁶. The Li⁺ transference number of SEs is unity; thus, practical Li⁺ conductivity is comparable to or even exceeds that of the organic electrolytes used in commercial lithium-ion batteries⁷.

The latter issue is related to the formation of low-resistivity and highly stable electrode–SE interfaces. Challenges in achieving low-resistivity interface begin with the interface bonding process, and their stability during charge‒discharge reactions depends on the potential window of the SEs^8,9 and (electro)chemical and/or mechanical factors^10,11. During the interface bonding process, resistive side reactions often occur around the interface^12,13,14,15 because both electrodes and SEs are composed of different elements and concentrations; this leads to mutual diffusion driven by the chemical potential differences of the constituent elements. A typical example is the case of SSBs using oxide-based SEs where high-temperature sintering is applied, and this process can facilitate the diffusion of various constituent elements. However, the most mobile element during the interface bonding process is Li because both materials are designed with Li⁺ conductive properties. When the Li chemical potential (${\mu }_{{{{\rm{Li}}}}}$) aligns at the interface, the Li concentration also varies around the interface. This nonstoichiometric Li region triggers side reactions that occur during the sintering process^16,17 and during the charging reactions of SSBs^18,19. Thus, understanding how the alignment of ${\mu }_{{{{\rm{Li}}}}}$ influences the low-resistivity electrode–SE interface is important.

Problems associated with the formation of a low-resistivity electrode–SE interface during the interface bonding process have been examined by using various electrode–SE combinations, and numerous approaches have been proposed, such as coating layer formation²⁰, in-situ interlayer formation^21,22, and low-temperature bonding processes^23,24. Among these approaches, several factors have been identified from the fundamental aspects of thin-film SSBs using amorphous lithium phosphorus oxynitride glass electrolytes (LiPONs, typically Li_3+xPO_4−xN_x)²⁵. The anodic potential window of LiPON has been estimated to be ca. 4 V in theoretical calculations^8,9. Thin-film SSBs typically employ Li metal anodes, and stable charge‒discharge reactions have been ensured for tens of thousands of cycles²⁶. The Li⁺ conductivity of LiPON is approximately 10⁻⁶ S cm^­¹ at room temperature, and LiPON is generally prepared by radio frequency (RF) magnetron sputtering. Factors contributing to high interfacial resistance during the formation of LiPON on electrodes have been discussed in terms of chemical potential differences²⁷, Fermi energy differences²⁸, and damaging effects during interface formation²⁹. Chemical potential differences are related to element diffusion during the interface bonding process, as mentioned above. For example, when LiCoO₂ (LCO) is combined with LiPON, both Li and O from LCO diffuse into LiPON, forming resistive lithium compounds such as LiNO₂ and reducing Co on the LCO side²⁷. Fermi energy differences pertain to the electron transfer; since the Fermi energy of LiPON is higher than that of LCO, electrons flow from LiPON to LCO upon bonding, reducing LCO²⁸. This leads to Co reduction and structural degradation of LCO. The damage layer effect involves the structural degradation of LCO during the deposition of LiPON. When LiPON is formed on an LCO film by sputtering under an N₂ atmosphere, ion bombardment with negatively charged N ions damages the LCO structure, and the interfacial resistance is subsequently increased²⁹. Consequently, the interfacial resistance of LCO/LiPON at 4.0 V (25 °C) can be reduced to as low as 8.6 Ω cm² by a less damaging sputtering position³⁰. All these factors are related to the reduction in the interfacial resistance. However, LiPON is an amorphous material with various Li compositions related to ${\mu }_{{Li}}$. Additionally, the Fermi energy varies depending on the Li/P atomic ratio³¹. Furthermore, the sputtering conditions, including their positions, differ from those used by other researchers. Thus, how each factor and its interactions contribute to the reduction in interfacial resistance remains unclear.

In this study, we prepared amorphous lithium phosphate thin films (LPOs) with low amounts of nitrogen (approximately 1% versus oxygen), which have different Li/P atomic ratios (2 to 9) on the film surface. The LPO films were fabricated using bias-induced RF magnetron sputtering at substrate temperatures less than ­80 °C. These LPO films were deposited onto c-axis oriented lithium cobalt oxide (LiCoO₂: LCO) thin-film electrodes, and their electrochemical properties were evaluated. We demonstrate that interface bonding is classified into two types, namely, lithium-insertion bonding and lithium-extraction bonding; the type depends on the Li/P atomic ratio. Furthermore, a strong correlation is observed between the Li/P atomic ratio and the LCO/LPO interfacial resistance. An optimal Li/P atomic ratio range achieves the lowest interfacial resistance of less than 10 Ω cm² at 4.0 V and 25 °C. Once the Li/P atomic ratio increases beyond this range, reductive degradation of LCO occurs because of excess Li insertion into LCO. Alternatively, when the Li/P atomic ratio decrease beyond this range, kinetically less reversible phase formation occurs because too much Li is extracted from LCO. Both these cases result in increased interfacial resistances.

Results and discussion

Figure 1a and b show the cross-sectional SEM images of the LPO films deposited under 0 V substrate bias on Pt/Ti/Si substrates at substrate temperatures less than −80 °C (CL film) or at room temperature (RT film). The film thicknesses of the CL and RT films were 1.7 μm and 1.5 μm, respectively, and did not significantly depend on the substrate temperature. However, the CL film exhibited uniform contrast, whereas the RT film contained granular contrasts within the film. Figure 1c shows the XRD patterns of the CL film, RT film, and substrate. The CL film showed a halo pattern at approximately 20°, in addition to the diffraction peaks from the substrate; however, the RT film contained other diffraction peaks (*), although they did not match the diffraction patterns from the typical impurity materials such as LiPO₃, Li₄P₂O₇, Li₃PO₄, Li₂CO₃, and Li₂O. These results indicate that the CL film is completely formed in an amorphous state, whereas the RT film is partially crystallized. Figure 1d and e show the electrochemical impedance spectroscopy (EIS) of the CL and RT films measured at 3.0 V using Li/LPO/Pt/Ti/Si cells (S-cells). Each spectrum exhibited a single semicircle in the higher frequency region and followed a divergent line to the real axis in the lower frequency region. The diameter of the semicircle is assigned to the resistance of Li⁺ conduction in the LPO films, and the Li⁺ conductivities at 25 °C (σ_Li+) were estimated to be 4.7 × 10⁻⁷ S cm⁻¹ for the CL film and 1.3 × 10⁻⁷ S cm⁻¹ for the RT film. The CL films are uniformly formed in amorphous states and have higher σ_Li+. Thus, we selected to prepare various LPO films under substrate cooling conditions in later discussions.

Fig. 1: Characterizations of LP-xs.

Cross-sectional SEM image of an LPO prepared at a cooling condition (CL) and b at room temperature (RT). c XRD patterns of LPOs prepared under CL and RT conditions in addition to reference peaks. EIS plots of the S-cells using LPOs prepared under d CL and at e RT conditions. f Summary of the oxygen chemical bonding state of LP-xs (x = 2.20, 2.63, 5.24, and 8.78) prepared under CL conditions; here, bridging oxygen (red): O_b, nonbridging oxygen (green): O_nb, and Li₂O (gray). g Ionic conductivities of LP-xs at 25 °C.

Supplementary Table 1 summarizes the Li/P, O/P, and N/O atomic ratios of the series of the LPO films prepared in this work. All LPO films were deposited at a substrate bias of 0 V to −6 V under substrate cooling (CL conditions). These atomic ratios were determined by the XPS measurements on each film surface. The Li/P atomic ratio varied from 1.75-8.78 depending on the substrate bias, whereas the N/O atomic ratio remained constant at 0.01–0.03. Both the Li/P and O/P atomic ratios tended to increase with increasing negative bias. Hereafter, the LPO films with a Li/P atomic ratio of x are denoted LP-x. Figure 1f shows the proportions of the bridging oxygen (O_b), nonbridging oxygen (O_nb), and Li₂O derived from O 1 s for LP-2.20, LP-2.63, LP-5.24, and LP-8.78. The increased Li/P atomic ratio resulted in an increase in the proportion of Li₂O³¹. Cross-sectional SEM images of these LP-xs (Supplementary Fig. 1a–d) showed uniform contrast as with Fig. 1a and their EISs measured at 3.0 V using S-cells (Supplementary Fig. 1e–h) were resembled with those in Fig. 1d. Figure 1g shows a summary of the σ_Li+ of the LP-xs. σ_Li+ was obtained in the range of 3-5 × 10⁻⁷ S cm⁻¹ regardless of x. These σ_Li+ values are approximately 1/7 to 1/4 of the typical σ_Li+ of LiPON (ca. 2 × 10⁻⁶ S cm⁻¹) because of the lower nitrogen substitution (N/P < 0.03) in our LPO films than in general LiPON (N/P~0.1)^32,33.

Figure 2a shows the open circuit voltages (OCVs) of the S-cells using the LP-2.20, LP-2.63, LP-5.24, and LP-8.78. OCV depends on the Li/P atomic ratio, and a higher Li/P atomic ratio tends to decrease the OCV. The OCV of the S-cell is the difference in ${\mu }_{{{{\rm{Li}}}}}$ at the Li/LP-x interface (${{{{\rm{\mu }}}}}_{{{{\rm{Li}}}}}^{{{{\rm{Li}}}}/{{{\rm{LP}}}}{{ \mbox{-} }}{{{\rm{x}}}}}$) and Pt/LP-x interface (${\mu }_{{{{\rm{Li}}}}}^{{{{\rm{Pt}}}}/{{{\rm{LP}}}}{{ \mbox{-} }}{{{\rm{x}}}}}$), as illustrated in the inset of Fig. 2a.

$${{{\rm{OCV}}}}=\frac{{{{{\rm{\mu }}}}}_{{{{\rm{Li}}}}}^{{{{\rm{Li}}}}/{{{\rm{LP}}}}{{ \mbox{-} }}{{{\rm{x}}}}}-{\mu }_{{{{\rm{Li}}}}}^{{{{\rm{Pt}}}}/{{{\rm{LP}}}}{{ \mbox{-} }}{{{\rm{x}}}}}}{F}$$

Fig. 2: Characterizations of the S-cells with LP-xs.

a OCV variations of the S-cells using the LP-xs (x = 2.20, 2.63, 5.24, and 8.78; 1–2 μm in thickness) and the L-cells using the LP-xs (x = 1.75, 2.60, and 4.23; 10 nm in thickness) at 25 °C. Inset shows the schematic image of the L-cell and observed OCV in the S-cell. b Schematic distribution images of ${\eta }_{{{{\rm{Li}}}}}$, ${\mu }_{{{{\rm{Li}}}}}$, and ${\varPhi }_{{{{\rm{Li}}}}}$ in the LP-xs during the deposition at substrate bias of 0 to −6 V against earth.

Here, the conduction band minimum (CBM) of LPO is lower than the work function of Li³¹. LPOs then react with Li to form a stable interface, similar to a LiPON film^15,24. Therefore, ${{{{\rm{\mu }}}}}_{{{{\rm{Li}}}}}^{{{{\rm{Li}}}}/{{{\rm{LP}}}}{{ \mbox{-} }}{{{\rm{x}}}}}$ is fixed regardless of the Li/P atomic ratio. The variations in the OCV are then reflected by the difference in ${\mu }_{{{{\rm{Li}}}}}^{{{{\rm{Pt}}}}/{{{\rm{LP}}}}{{ \mbox{-} }}{{{\rm{x}}}}}$ across each cell. Bias-induced RF sputtering has been applied to alter the atomic ratios of LiPON films^34,35, and a mismatch of the Li/P atomic ratios at the surface and inside the bulk has been reported³⁵. We measured the average Li/P atomic ratios of LP-2.20, 2.86, and 5.73 using the inductively coupled plasma (ICP) and atomic absorption spectrometry, and the Li/P ratios were determined to be 2.78-2.79 in these films. Additionally, we measured the depth profiles of both the Li/P and O/P atomic ratios in those films using nuclear reaction analysis (NRA) and Rutherford backscattering spectrometry (RBS), and the values were 2.80–2.81 and 3.15, respectively, with uniform distributions throughout the films (Supplementary Fig. 2). These results indicate that the surface composition varies with the substrate voltage, whereas the bulk composition is similar in these films. This situation reasonably explains the attainment of nearly the same ${\sigma }_{{{{{\rm{Li}}}}}^{+}}$ for the LP-xs samples regardless of the Li/P atomic ratio (Fig. 1g). Here, to determine the relationship between the Li/P atomic ratio around the Pt/LP-xs and ${\mu }_{{{{\rm{Li}}}}}^{{{{\rm{Pt}}}}/{{{\rm{LP}}}}{{ \mbox{-} }}{{{\rm{x}}}}}$, we prepared 10 nm-thick LP-1.75, 2.60, and 4.23 layers on the Pt/Ti/Si substrates as quasi-uniform interface layers around the Pt/LP-xs. We then measured their OCVs through an organic liquid electrolyte, with the Li metals as reference and counter electrodes (L-cells). As shown in Fig. 2a, the OCV of the L-cells almost followed the variations in the S-cells. These results indicate that the Li/P atomic ratio around the Pt/LP-xs changes in relation to the surface Li/P atomic ratio.

A schematic image of LP-x prepared at different substrate biases is shown in Fig. 2b. When the substrate bias is set to a negative voltage, Li is preferably deposited on the LP-x surface, increasing the Li/P atomic ratio at the LP-x surface. LP-xs maintain uniform Li/P and P/O atomic ratios in the bulk, whereas the Li/P atomic ratio around the Pt/LP-x interface varies in relation to the surface Li/P atomic ratio. We speculate this behaviour occur by following mechanism. Both the LP-x surface and substrate holder are exposed to plasma with electron conduction during sputtering. Since LP-xs are Li⁺ conductive materials with negligible electronic conductivity³⁶, this short-circuited state aligns ${\mu }_{{{{\rm{Li}}}}}^{{{{\rm{Pt}}}}/{{{\rm{LP}}}}{{ \mbox{-} }}{{{\rm{x}}}}}$ with ${\mu }_{{{{\rm{Li}}}}}$ on the LP-xs surface during the deposition of LP-xs and will result in a total alignment of the electrochemical Li⁺ potential (${\eta }_{{{{\rm{Li}}}}}$). However, further experimental studies are needed to support this mechanism.

Figure 3a–j show the cyclic voltammograms (CVs) at 25 °C of thin-film batteries (TFB; Li/LP-x/LCO; 1.98 ≤ x ≤ 8.78). The OCV of each TFB is indicated by a red circle in each CV. The CVs were initially swept from the OCV to 3.0 V (OCV-sweep), followed by three cycles at 3.0–4.0 V (1st-3rd CVs). Two cycles were subsequently conducted at 3.0–4.2 V (4th-5th CVs). Both the 1st CV and the 4th CV overlapped with the 2nd CV and the 5th CV, respectively, with the exception of the TFBs using the LP-5.73 and LP-8.78. For the TFBs with LP-x values of 2.27 ≤ x ≤ 3.52 (Fig. 3c–g), a couple of redox peak due to the two-phase reaction of LCO at 3.9 V and two pairs of redox peaks at 4.1 V and 4.2 V corresponding to the order‒disorder phase transition was observed³⁷. For the TFB with LP-5.24 (Fig. 3h), an additional oxidation peak appeared at 2.0–2.7 V during the OCV-sweep, indicating the formation of spinel-phase LCO^38,39. For the TFBs with LP-x values of 5.73 ≤ x (Fig. 3h–j), the spinel-phase oxidation peak further increased with increasing x, whereas the two-phase reaction peak decreased and became broader. On the other hand, for the TFBs with LP-x values of x ≤ 2.20 (Fig. 3a, b), the two-phase reaction peak weakened with decreasing x. These results clarify that the charge‒discharge reactions of the TFBs are strongly influenced by the range of x. Figure 3k–t show the EISs of the TFBs (Li/LP-x/LCO; 1.98 ≤ x ≤ 8.78) measured at 3.0 V after OCV-sweep and at 4.0 V during the 3rd CV. In the TFBs with LP-x values of 2.27 ≤ x ≤ 3.52 (Fig. 3m–q), one semicircular arc was observed in the higher frequency region at both voltages, and the interfacial resistance at 4.0 V (R_int) was too small to detect as a separated semicircular arc in the middle frequency region (Supplementary Fig. 3). In the TFB with LP-5.24 (Fig. 3r), R_int was distinguished in the middle frequency region and further increased in the TFBs with LP-5.73 (Fig. 3s) and LP-8.78 (Fig. 3t). Additionally, in the TFBs with LP-2.20 (Fig. 3l) and LP-1.98 (Fig. 3k), a semicircular arc visibly appeared in the middle-frequency region assigned to R_int.

Fig. 3: Electrochemical properties of the TFBs depending on the LP-xs.

The variations of CV (a–j) and EIS (k–t) of the TFBs using LP-1.98 (a and k), LP-2.20 (b and l), LP-2.27(c and m), LP-2.40(d and n), LP-2.52 (e and o), LP-2.86 (f and p), LP-3.52 (g and q), LP-5.24 (h and r), LP-5.73 (i and s), and LP-8.78 (j and t) measured at 25 °C. Each TFB shows a different OCV, which is shown as a red circle in each CV. OCV-sweep (black) means the initial CV from the OCV to 3.0 V. CVs were then repeated at 3.0–4.0 V (red) for two cycles, and the EIS was measured at the 3rd cycle. Later, two cycles were carried out at 3.0–4.2 V (blue). The 1st CV is denoted as 1 in the CV using the LP-5.73 and LP-8.78. EISs were measured at 3.0 V (black) after the OCV-sweep and 4.0 V (red) during the 3rd CV. In the cases of TFBs using LP-5.73 and LP-8.78, EISs were also measured at 2.5 V during the OCV-sweep to check the operations. The numbers in the EIS plots are the data measured at 10ⁿ Hz. u Summary of the OCVs and the R_int depending on the LP-xs.

Figure 3u show a summary of both the OCV and the R_int against x in LP-x for the TFBs. The pristine LCO film showed two Raman bands (A_1g and E_g) characteristic of the layered structure of LCO⁴⁰, and both the OCV and the R_int were 3.33 V (vs. Li/Li⁺) and 110 Ω cm² in an organic liquid electrolyte, respectively (Supplementary Fig. 4). On the other hand, the OCVs of the TFBs depended on x and tended to decrease with increasing x; these results were similar to those using the S-cells. The estimated R_int values of the TFBs obtained via LP-x (2.27 ≤ x ≤ 3.52) were less than 10 Ω cm², with value of 1.4 Ω cm² (LP-2.27), 5.7 Ω cm² (LP-2.40), 1.4 Ω cm² (LP-2.52), 4.3 Ω cm² (LP-2.86), and 7.1 Ω cm² (LP-3.52), as shown in the Supplementary Fig. 3. The R_int for the TFB using the LP-5.24 increased to 140 Ω cm² and further increased to thousand Ω cm² for the TFBs using the LP-5.73 and LP-8.78. Additionally, the R_int for the TFB using the LP-2.20 and LP-1.98 exceeded 800 Ω cm². Supplementary Table 2 summarizes both the OCVs and the R_ints against LP-xs. The R_int reached a minimum in the range of 2.27 ≤ x ≤ 3.52 and then increased once the xs exceeded this range.

Figure 4a shows the XRD patterns of LP-1.98, LP-2.40, LP-2.52, LP-2.86, LP-3.52, and LP-5.73 deposited on the LCO films. The XRD patterns of the LCO film without LP-x (bare LCO) and the Pt/Ti/Si substrate are also shown as references. Both the peak intensities and diffraction angles are normalized using Pt (111) at 2θ = 39.90° from each substrate. The diffraction peaks from the bare LCO film were observed at 18.97° for (003) and 38.50° for (006), and the c-axis lattice constant (c_axis) was determined to be 14.03 Å nearly consistent with a previous report of LCO⁴⁰. Figure 4b shows a magnified image around the diffraction peak at 19° in Fig. 4a. The LCO film covered with LP-2.86 had the same c_axis of 14.03 Å with maintaining (003) diffraction intensity of the bare LCO film. The LCO film covered with LP-x (3.52 ≤ x) decreased (003) diffraction intensity with increasing x. Alternatively, the LCO films covered with LP-x (x ≤ 2.52) shifted the (003) peak depending on x. The LCO with LP-2.40 (OCV in TFB; 4.37 V) shifted (003) peak to a lower angle, and the c_axis of the LCO film expanded to 14.36 Å; this value was nearly consistent with the expected c_axis at 4.4 V for LCO⁴¹. Additionally, the LCO with LP-1.98 (OCV in TFB; 4.62 V) shifted (003) peak to higher angle, and the c_axis shrink to 13.62 Å; this value was in agreement with expected c_axis at 4.6 V for LCO⁴². These results clarify that the LCO structure changes depending on x during the interface bonding process.

Fig. 4: Characterizations of the LCO films after depositing LP-xs.

a XRD patterns of a pristine LCO film (orange) and the LCO films after depositing LP-1.98 (gray), LP-2.40 (light blue), LP-2.52 (brown), LP-2.86 (blue), LP-3.52 (green), and LP-5.73 (purple). b Magnified XRD pattern (17-22°) of a. Cross-sectional STEM images of LCO/LP-xs (c–f) and Co L-edge EELS spectra measured at A-D areas in STEM (g–j) for bare LCO (c and g), LCO/LP-3.52 (d and h), LCO/LP-2.40 (e and i), and LCO/LP-5.73 (f and j). k Summary of Co L3-edge energy loss measured at A-D areas in bare LCO (black), LCO/LP-3.52 (green), LCO/LP-5.73 (blue), and LCO/LP-2.40 (red).

To further investigate the structural changes in the LCO films covered with LP-xs, STEM-EELS measurements were performed. Figure 4c–f show the cross-sectional STEM images of the bare LCO (Fig. 4c) and the LCO films covered with LP-3.52 (Fig. 4d), LP-2.40 (Fig. 4e), and LP-5.73 (Fig. 4f). EELS spectra around the Co L-edge (770-800 eV; Fig. 4g–j) and O K-edge (520–550 eV; Supplementary Fig. 5a–d) were measured in the A-D areas of each STEM image for the bare LCO film (Fig. 4g and Supplementary Fig. 5a) and LCO films covered with LP-3.52 (Fig. 4h and Supplementary Fig. 5b), LP-2.40 (Fig. 4i and Supplementary Fig. 5c), and LP-5.73 (Fig. 4j and Supplementary Fig. 5d). Here, the oxidation states inside LCO were estimated based on the energy loss of the Co L2-edge (~795 eV) and Co L3-edge (~780 eV) peaks. The values of the Co L3-edge energy in the A-D areas inside the LCO films are plotted in Fig. 4k. The Co L3-edge shifted to lower energy with increasing x, indicating that Co is reduced state in this order, and this trend explains decreasing OCV variations. The oxidation states of the LCO film were also estimated from the pre-edge peak of the O K-edge (~530 eV); this corresponds to the transition of O 1s states to O 2p states hybridized with Co 3d states (Supplementary Fig. 5). The pre-edge peak intensity is related to the oxidation state, and the pre-edge peak intensity from LCO with LP-2.40 is stronger than those from bare LCO and LCO with LP-3.52; these results indicate that LCO with LP-2.40 is in a more oxidized state⁴³. In the case of LCO with LP-2.40, Co L3-edge peak energy is flat inside the film, indicating that delithiated LCO is uniformly formed in the film. Because this Li extraction corresponds to the charging reaction of LCO within an acceptable voltage range and then proceeds smoothly, Li is extracted evenly across the film. On the other hand, in the case of LCO with LP-5.73, the pre-edge peak of the O K-edge gradually decreased from the substrate to the surface, and the pre-edge peak intensity nearly disappeared both in the D and C area (ca. 20 nm away from the interface); these results indicate that LCO has a reductively degraded region from the interface to the substrate⁴⁴. Excess Li insertion induces an irreversible structural transformation, and such transformation rate will be much slower than that of Li diffusion within an acceptable voltage range. This provides the formation of gradually degraded region from the interface. In the case of LCO with LP-3.52, the Co L3-edge showed a lower energy shift near to the interface (D position) compared with the bulk, but this trend is also observed in the bare LCO. Thus, LCO film is totally reduced state, but the interface structure will be almost maintained as with the bare LCO.

Figure 5 shows a schematic image of the LCO/LP-x interface during interface bonding depending on x in LP-x. Once the LCO films are combined with LP-xs, the LCO films are formed at charged or discharged states depending on x, and provide different OCVs. This situation arises from the alignment of ${\mu }_{{{{\rm{Li}}}}}$ at the LCO/LP-x interface during interface bonding. The LP-x with a lower ${\mu }_{{{{\rm{Li}}}}}$ around the interface aligns with LCO by extracting Li from the LCO film, whereas the LP-xs with a higher ${\mu }_{{{{\rm{Li}}}}}$ around the interface align with LCO by inserting Li into the LCO film. When these Li extracted or inserted states of LCO are within an acceptable range (2.27 ≤ x ≤ 3.52), the LCO/LP-xs interfaces are formed in a low-resistive interface state. When x becomes greater than the acceptable range, extra Li insertion reductively decomposes LCO as suggested by a weaker (003) diffraction peak and forms a resistive spinel phase³⁸. This reduced region preferentially forms around the interface according to the EELS analyses (Fig. 4). On the other hand, when x becomes lower than the acceptable range, too much Li is extracted from LCO, and a kinetically less reversible phase transition (H1-3/O1 phase) occurs, and another resistive region forms⁴⁵. LCO with LP-1.98 has a broad shoulder X-ray diffraction peak at approximately 20.8° (c_axis~13.00 Å), and this peak is basically consistent with perfectly delithiated CoO₂ (c_axis = 12.88 Å); consequently, the Li-extracted region will also form a gradient state in the thickness direction. Li-extraction type bonding has been observed in the case of LCO-Li_1+xAl_xTi_2−x(PO₄)₃¹⁶, and Li-insertion type bonding has been reported in the case of Li_0.5CoO₂-Li_6.4La₃Zr_1.4Ta_0.6O₁₂¹⁷. Thus, systematic variations in LCO-LP-x depending on x are not special cases but are common phenomena in all-solid-state batteries at the interface bonding process depending on the difference in ${\mu }_{{{{\rm{Li}}}}}$ between the electrode and SE.

Fig. 5: Schematic image of the two-types of interface bonding situations around LCO/LP-xs.

The OCVs of the TFBs depend on the LP-xs and LP-2.86 shows nearly the same OCV as that measured in liquid electrolyte. Then, this ${\mu }_{{{{\rm{Li}}}}}$ is nearly the same as that of pristine LCO. Once the x decreases below 2.86, ${\mu }_{{{{\rm{Li}}}}}$ in the LP-xs becomes lower than that of LCO; here, the interface is then formed with the Li-extraction, and the charged LCO is also formed. Highly Li-extracted LCO provides a kinetically less reversible H1-3/O1 phase or decomposes the structure, leading to an increase of R_int. On the other hand, when x becomes larger than 2.86, ${\mu }_{{{{\rm{Li}}}}}$ in the LP-xs becomes higher than that of LCO; here, an interface is formed as Li-insertion states, and discharged LCOs are also formed. Highly Li-inserted states (5.73 ≤ x) reductively decompose the LCO structure, leading to the increase of R_int.

The alignment of ${\mu }_{{{{\rm{Li}}}}}$ at the LCO/LP-x interface occurs during the charge‒discharge reactions. Thus, we can reproduce the above degradations by electrochemical methods. Here, we selected the TFBs using LP-2.27 or LP-3.52 with R_int values less than 10 Ω cm² and investigated the effects of the lower or higher cut-off voltages on the CVs (Fig. 6a, b, g, and h) and EISs (Fig. 6c–f and i–l). Figure 6a and b show the CVs of a TFB using LP-2.27 at 1.2–4.0 V and at 1.0–4.0 V, respectively. The voltage was initially swept from the OCV (4.01 V) to 3.0 V, and then the voltage was swept from 3.0 V to 1.2 V, 4.0 V, and 3.0 V at the 1st CV. Then, the 2nd CV was carried out at 3.0–4.0 V for EIS measurements at 3.0 V and 4.0 V. The 3rd CV was carried out as with the 1st CV and held at 1.2 V for 15 h, and then the EISs were measured at 3.0 V and 4.0 V during the 4th CV as with the 2nd CV. Later, the same measurements were conducted by changing the lowest cut-off voltage from 1.2 V to 1.0 V. The CVs measured at 1.2–4.0 V (Fig. 6a) showed two-phase reaction peaks at 3.9 V, and the corresponding EISs at 3.0 V and 4.0 V at the 2nd and the 4th CVs (Fig. 6c and d) almost overlapped with small R_int, as shown in Fig. 3m–q. When CVs were conducted by changing the lowest cut-off voltage to 1.0 V (Fig. 6b), an irreversible reductive current flowed at 1.0 V in the 1st CV and generated oxidation peaks at 2.0–2.7 V, whereas the two-phase reaction peak decreased. The EIS at 4.0 V measured during the 2nd CV (Fig. 6e) and the 4th CV (Fig. 6f) generated a tilted capacitive line to the real axis in the low-frequency region. These characteristic variations in the CVs and EISs are observed in the TFBs using LP-5.73 (Fig. 3i and s) and LP-8.78 (Fig. 3j and t). The effects of higher cut-off voltages on the electrochemical performance of a TFB were investigated in the same manner. Figure 6g and h show the CVs of a TFB using LP-3.52 at 3.0–5.0 V and at 3.0–5.2 V, respectively. The voltage was initially swept from the OCV (1.78 V) to 3.0 V, and then the CVs were measured at 3.0–5.0 V at the 1st CV. Then, the 2nd CV was carried out at 3.0–4.0 V for the EIS measurements at 3.0 V and 4.0 V. The 3rd CV was carried out as with the 1st CV and held at 5.0 V for 15 h, and then the EIS was measured at 3.0 V and 4.0 V during the 4th CV. Later, the same measurements were conducted by changing the higher cut-off voltage from 5.0 V to 5.2 V. The 1st CV measured at 3.0–5.0 V (Fig. 6g) showed almost reversible redox peaks due to the two-phase reaction at 3.9 V, the order-disorder reaction at 4.1–4.2 V, the O3/H1-3 phase transition reaction at 4.5 V, and the H1-3/O1 phase transition reaction at 4.6 V⁴⁴. The 3rd CV peak appeared as a broad two-phase reaction peak during the cathodic scan, and both the O3/H1-3 and H1-3/O1 reaction peaks also disappeared. The R_int measured at the 2nd CV (Fig. 6i) was estimated to be 7.1 Ω cm²; this value was slightly increased to 20 Ω cm² at the 4th CV (Fig. 6j). When the cut-off voltage increased to 5.2 V, the two-phase reaction current peak further decreased and broadened (Fig. 6h), and R_int increased to 128 Ω cm² at the 2nd CV (Fig. 6k) and further to 1280 Ω cm² at the 4th CV (Fig. 6l). A broad reductive current at approximately 3.9 V (Fig. 6h) was observed at the OCV-sweep of the TFB using LP-1.95 (Fig. 3a) and LP-2.20 (Fig. 3b), and R_int visibly increased in the same manner (Fig. 3k and l). At these higher cut-off voltages, R_int increase because of kinetically less reversible H1-3/O1 phase formation⁴⁵.

Fig. 6: Electrochemical degradations of the TFBs depending on the cut-off voltages.

CVs (a and b) and EISs (c–f) of TFBs using LP-2.27 were measured from 1.2 V (or 1.0 V) to 4.0 V. After the OCV-sweep, the 1st CV (black) was scanned between 1.2 V and 4.0 V. EIS measurements (Fig. 4c) were then taken at 3.0 V (black) and 4.0 V (red) during the 2nd CV. The 3rd CV repeated the 1st scan, but the voltage was held at 1.2 V for 15 hours. Subsequently, EISs (Fig. 4d) were measured at 3.0 V (black) and 4.0 V (red) during the 4th CV. After this series of electrochemical measurements, additional CVs (b) and EISs (e during the 2nd CV and f during the 4th CV) were similarly conducted, with the lower cut-off voltage changed to 1.0 V. To investigate the effects of a higher cut-off voltage, CVs (g and h) and EISs (i–l) of TFBs using LP-3.52 were measured in the same manner at 3.0V-5.0 V (g and i during the 2nd CV, and j during the 4th CV) and at 3.0V–5.2 V (h and k during the 2nd CV, and l during the 4th CV). All CVs were conducted at a scan rate of 0.5 mV s⁻¹, and EIS measurements ranged from 100 mHz to 200 kHz. The numbers in all the EIS are data points measured at 10ⁿ Hz. All electrochemical measurements were performed at 25 °C.

The reductive decomposition of LCO during the interface bonding process with LiPON has been reported to occur by electron transfer from LiPON to LCO, which is driven by the difference in the Fermi energy (ΔE_F) between LCO and LiPON²⁸. We investigated the band structure of LP-xs (x = 2.35, 2.82, 3.77, and 4.05) using UPS and LEIPS³¹. The VBM depends on the Li/P atomic ratio, where the VBMs are approximately 8.8 eV for Li/P atomic ratios less than 3, and the values decrease to 7.0 eV for Li/P atomic ratios greater than 3. The CBM is nearly independent of the Li/P atomic ratio, and the values are 3.4–3.7 eV. Here, the E_F of LP-2.35 is 4.5 eV greater than that of LCO (5.3 eV)⁴⁶, and that of LP-3.77 is 5.3 eV comparable with that of LCO. From the difference in ΔE_F, LCOs are expected to be reduced by the combination with LP-2.35 and to be nearly unchanged by the combination with LP-3.77. However, LP-2.35 is expected to oxidize LCO, whereas LP-3.77 is expected to reduce LCO during the bonding process, as shown in Fig. 3u. Thus, the difference in ΔE_F is not sufficient to explain the trend of the oxidative/reductive trends of LCO depending on x. Sputtering damage to the LCO structure during the deposition of LPO has been reported to influence R_int²⁹. Our LP-xs are prepared with the same sputtering dimensions (on-axis), but the values of R_int are strongly related to the Li/P ratios. Thus, the sputtering damage is not directly related to the degradation of LCO in our systems. However, sputtering damage may influence the resistance in another way. We prepared LP-2.79 on an LCO film and a Pt/Ti/Si substrate at RT (LP-2.79 (RT)), where the LP-x is within the range that provides a low-resistive R_int, as shown in Fig. 3u. The OCV of TFB using LP-2.79 (RT) was 4.40 V; this value was higher than that prepared under substrate cooling conditions. The CV plot showed the typical redox peaks of LCO (Supplementary Fig. 6a). The EIS data of the S-cell of Li/LP-2.79/Pt/Ti/Si (OCV: 3.01 V) revealed one semicircular arc at 3.0 V, as shown in Fig. 1d. On the other hand, the TFB using LP-2.79 (RT) presented an extra semicircular arc in the higher frequency region (Supplementary Fig. 6b), whereas R_int was low, and the EIS data measured at 3.0 V and 4.0 V nearly overlapped. Because the higher-frequency semicircular arc observed in TFB was independent of the voltage, we speculate that some Co²⁺ may diffuse into the LPO because of the effect of sputtering damage, and that a smaller Li⁺ conductive region is formed because of the reduced Li⁺ carrier density around the LCO/LP-2.79 (RT) interface. The structural stability of charged or discharged LCO depends on the temperature. Thus, when the substrate is not sufficiently cooled, the thermal decomposition of Li-extracted or Li-inserted LCO will provide other resistive compounds depending on the preparation conditions³⁰.

Finally, we investigated the cycle stability of TFB using LP-2.27 for 10000 cycles at 3.0–4.2 V and 25 °C at 100 C rate. Supplementary Fig. 7a shows the charge‒discharge curves of the TFB. The TFB delivered 2.30 μAh cm⁻² during the initial discharge reaction, and the capacity was basically maintained after 10000 cycles, with a degradation ratio per cycle of 0.002% (Supplementary Fig. 7b). Although the initial discharge capacity at 100 C was ca. 63% of the theoretical capacity of the LCO film (ca. 50 nm in thickness, 3.8 μAh cm⁻²), this capacity increased to 3.32 μAh cm⁻² at 1 C rate and accounted for ca. 87% of the theoretical capacity (Supplementary Fig. 7a). This means that our LCO film is an c-axis oriented film, and Li diffusion inside the LCO film is retarded at higher C rates, which is different from that observed on epitaxial LCO films⁴⁷. EISs were measured after the 1st, 1000th, 2000th, 5000th, and 10,000th cycles (Supplementary Fig. 7c–g). Each EIS showed one semicircular arc in the high-frequency region followed by a straight line to the real axis. The R_int slightly increased from 1.4 Ω cm² (Supplementary Fig. 7h) to 15.6 Ω cm² (Supplementary Fig. 7i) after 10,000 cycles, and a stable interface was maintained. Of course, these stabilities will depend also on other factors such as mechanical strength, higher temperature aging, which will be addressed in future work.

Conclusions

LCO thin films with c-axis orientations were combined with amorphous lithium phosphate glass electrolyte films, LPO, with different ${\mu }_{{{{\rm{Li}}}}}$ around the LCO/LPO interfaces. The correlations among the surface Li/P ratio, open-circuit voltage (OCV), and interfacial resistance between LCO and LPO were investigated. The ${\mu }_{{{{\rm{Li}}}}}$ is aligned during the bonding of LCO with LPO, and the bonding type is classified into two categories: Li-insertion type and Li-extraction type. An optimal range of Li/P ratios provide acceptable Li-inserted or Li-extracted states of LCO and maintain the crystal structure of LCO; this leads to a reduced interfacial resistance as low as 1.4 Ω cm². Once the Li/P atomic ratio exceed this optimal range, excess lithiated or delithiated LCOs are generated. These situations cause reductive decomposition of LCO or kinetically less reversible phase transitions (H1-3/O1 phase), leading to an increase in the interfacial resistance. These situations were reproduced using electrochemical overcharging or over-discharging states. During the electrode-solid electrolyte interface bonding process, the difference in ${\mu }_{{{{\rm{Li}}}}}$ plays an important role to supress side reactions. Thus, appropriate control of ${\mu }_{{{{\rm{Li}}}}}$ and/or desirable electrode‒solid electrolyte combinations is crucial for suppressing the side reactions during the interface bonding process. The findings of this study could be extended to other electrode materials, such as LiFePO₄ and LiNi_xCo_yMn_zO₂ (x + y + z = 1). However, because the thermal and structural stability of electrodes in their lithiated and/or delithiated states depend on each electrode material, optimized ranges will also depend on the electrode species.

Methods

Materials synthesis

The substrates used in this study were prepared by initially depositing a Ti thin film (thickness (t)~2 nm) and then a Pt thin film (t~60 nm) as a current collector via radiofrequency (RF) magnetron sputtering onto the entire surface of a Si(001) wafer (10 mm in square, t = 0.5 mm) with a naturally oxidized thin layer (SiO₂). The resulting substrates are denoted as Pt/Ti/Si substrates. Polycrystalline LiCoO₂ (LCO) thin films (ϕ = 6 mm, t~50 nm) were prepared on the Pt/Ti/Si substrates using pulsed laser deposition (PLD). The laser source was a 4th harmonic YAG laser, and the laser was focused on a Li_1.4CoO_x sintered target (Toshima Manufacturing Co., Ltd., Japan) with a laser energy of 0.25 J cm⁻² on the target at 10 Hz. The LCO films were fabricated at 600 °C on Pt/Ti/Si substrates under an O₂ partial pressure of 27 Pa for 1 h. The resulting LCO films were transferred to an RF magnetron sputtering system without being exposed to air. A series of LP-x thin films were deposited on the LCO films on Pt/Ti/Si substrates using bias-controlled RF magnetron sputtering with a substate cooling system³¹. A commercially available sintered Li₃PO₄ target (ϕ = 75 mm, t = 5 mm; Kojundo Kagaku, Japan) was used as a target, and indium was used as a bonding metal. The substrate temperature was decreased to less than ­80 °C during the deposition, where liquid nitrogen was added to the upper side of the substrate holder. The LP-x films were deposited at 120 W under a N₂ + O₂ (4:1 in volume) partial pressure of 1.5 Pa for 15 h. The resulting multilayers were then transferred into an Ar-filled glove box using a transfer vessel. The Li films (ϕ = 3 mm, t~10 μm) were deposited on the LP-x/LCO multilayers to form TFBs (Li/LP-x/LCO/Pt/Ti/Si) or S-cells (Li/LP-x/Pt/Ti/Si) using vacuum evaporation and assembled in an Ar-filled glove box.

Characterization methods for the LPO films

XPS and FE-SEM

Before depositing Li, the S-cells were transferred from the abovementioned Ar-filled groove box to the XPS instrument via an airtight sample holder. XPS measurements of the Li 1s, P 2p, O 1s, C 1s, N 1s, and Pt 4f peaks were collected using an XPS instrument (PHI5000 Versa Probe III, ULVAC-PHI). The X-ray source (Al Kα:1486.6 eV) was radiated to a sample surface with a spot size of 100 μm in square, a take-off angle of 80°, and a pass energy of 27 eV. The depth resolution is a few nm. All LP-x compositions described in this study were calculated using PHI MultiPak, a data analysis software installed in PHI5000. The Shirley method was used for baseline correction, and each spectrum was fitted using a Gaussian-Lorentzian function. The area of each fitted peak was then divided by corrected relative sensitivity factors of each element for the equipment provided within the software to determine the elemental composition. Each O 1 s spectrum was deconvoluted using CASA XPS software.

Both the thickness and morphology of the LP-xs were observed by field emission-scanning electron microscopy (FE-SEM; SU8030, Hitachi). Initially, a thin Au film was deposited on the surface of the samples via RF magnetron sputtering at room temperature without exposing the samples to air. The gold-coated samples were transferred into FE-SEM by an airtight sample holder. The FE‒SEM images were collected at an accelerating voltage of 10 kV.

RBS/NRA & ICP‒AES

The element distributions of P and O in addition to those of C, Si, Pt, Ti, and Cl in the LP-xs on Pt/Ti/Si were analysed by Retherford backscattering spectroscopy (RBS, National Electrostatics Corporation, Pelletron at Toray Research Centre, Inc.), and the element distribution of Li was analysed using nuclear reaction analysis (NRA). RBS was measured using a high-resolution RBS analyser; the ion beam current and size on the sample were approximately 13 nA and ϕ = 2 mm, respectively. The same apparatus was used for NRA; here, NRA was used to analyse the depth of the Li composition. The ⁷Li (p, α)⁴He reaction was used for the NRA measurements, with H⁺ (1500 keV) as the incident ion, and the beam current and current on the sample were 3 nA and ϕ = 2 mm, respectively. The average bulk compositions of the LP-xs were estimated using atomic absorption spectrometry (AAS; ZA 3300, Hitachi) and inductively coupled plasma spectrometry (ICP; PS3520VDDII, Hitachi). The samples measured by RBS and NRA were dissolved in dilute nitric acid and then distilled with water to prepare sample solutions. After the thin film was dissolved, the substrates were dissolved in sulfuric acid, nitric acid, hydrochloric acid, and hydrofluoric acid and then dissolved in diluted nitric acid to obtain a sample solution. The Li composition of these solutions was measured via AAS, the P composition was measured via ICP, and the amount of each element in each sample was determined and then converted to atomic ratios.

Characterization methods for the LCO films

XRD and Raman spectroscopy

The crystal structures of the LCO thin films were analysed using X-ray diffraction (XRD: Ultima IV, Rigaku) with a CuKα source. Scan mode of the XRD measurements was carried out at θ-2θ scans in air at room temperature in the range of 2θ = 15° to 60° with a sampling width of 0.01° and a scan speed of 2° min⁻¹. Both the diffraction peak intensities and angles were normalized to the Pt (111) peak intensity and diffraction position observed at 2θ = 39.90°. Raman spectroscopy (NRS-5100M, Jasco) was used to analyse a pristine LCO film in air at room temperature with a green laser (532 nm in wavelength).

STEM-EELS

The sample thinning for the STEM observations was carried out using a focused ion beam (FIB) coupled with SEM (FIB-SEM, NB5000, Hitachi High-Tech). A Ga-ion beam was radiated to the samples at an acceleration voltage of 40 kV at −160 °C, and the damage layer was removed at 10–5 kV using Ar gas. The electron energy-loss spectra (EELS) were collected using a 200 kV transmission electron microscope (JEM-2400FCS, JEOL Ltd.) equipped with a Gatan Continuum ERS imaging filter (Gatan Inc.). The acceleration voltage, probe current, pixel size, and resolution were 200 kV, 46 pA, 2 nm/pixel, and 48 × 40 pixels, respectively. The resulting spectral energy positions were normalized to zero-loss energy, and their intensities were normalized to the peak intensity of the Co L3-edge (~780 eV) and the pre-edge peak of the O K-edge (~530 eV) measured in the A area in Fig. 4c–f, respectively.

Electrochemical measurements

Electrochemical measurements of the TFBs were carried out via a potentio-galvanostat (VMP3, Biologic) in an Ar-filled glove box at 25 °C, where the TFBs were pressed ca. a few tens of kPa to prevent mechanical short circuit. CV measurements were carried out at 0.5 mV s⁻¹ in any case. EIS measurements were carried out with an alternating current (AC) amplitude of 10 mV at a given voltage from 100 MHz to 200 kHz. In both cells, the smallest electrode area is Li metal; thus, both the current density (mA cm^­2) in the CVs and the interface resistivity (R_int, Ω cm²) were calculated from the electrode area of Li metal (ϕ = 3 mm, 0.071 cm²). For comparison, the electrochemical properties of a pristine LCO film were measured in an organic liquid electrolyte. A three-electrode cell was fabricated in an Ar-filled glove box using an LCO film as the working electrode, Li metal as the reference and counter electrodes, and 1 mol dm⁻³ LiClO₄ dissolved in propylene carbonate as the organic liquid electrolyte.

Charge‒discharge reactions of TFBs using LP-2.27 were carried out by constant current‒constant voltage methods at a 1 C rate for 5 cycles and at a 100 C rate for 10,000 cycles. The reactions were conducted at 25 °C between 3.0 V and 4.2 V at 3.38 μA cm⁻² for the 1 C rate and 338 μA cm⁻² for the 100 C rate, where the TFBs were maintained at each cut-off voltage for 10 minutes. All these measurements were carried out in an Ar-filled glove box.