Operando and ex situ Raman spectroscopies for evaluating carbon structural changes in anode-free-type sulfide-based all-solid-state Li-ion batteries

Abstract

Anode-free-type sulfide-based all-solid-state Li-ion batteries are promising candidates for achieving high energy density and cycle capability. Although the introduction of an Ag/C layer between the solid electrolyte and the current collector could further improve the cycle capability, its role during charge–discharge remains unclear. In this study, we applied operando/ex situ Raman spectroscopies to observe structural changes in the Ag/C layer. Operando Raman spectroscopy revealed significant changes in the peak shape, intensity ratio, and position of the D- and G-bands derived from carbon black, suggesting Li intercalation into carbon even though it exhibited an amorphous structure. Additionally, new peaks appeared and disappeared during charging and discharging, respectively, corresponding to the deposition and dissolution of metallic Li onto the current collector. In contrast, ex situ Raman spectroscopy exhibited gradual changes in the D- and G-bands despite the clear formation of a metallic Li deposition layer. Therefore, we discovered that amorphous carbon black undergoes Li intercalation during charging and rapid structural changes under dynamic (operando) conditions compared with static (ex situ) conditions.

---

Introduction

Recently, renewable energies such as wind, solar, and hydropower have attracted significant attention as alternatives to fossil fuels to lower carbon emissions.¹ Although such renewable power sources usually fluctuate according to circumstances such as weather, a constant power supply is possible using a stationary power supply system with a large secondary battery, which stores a large amount of energy and offers high efficiency. Accordingly, electric vehicles (EVs), in which large Li-ion batteries (LIBs) are equipped, are anticipated not only to substitute combustible engines but also for storing large amount of energy for household purposes.^2–8

For such battery applications, energy density and safety during charging are particularly important because the former directly affects the driving range of the EV, and the latter is indispensable to prevent serious fires or explosions.

All-solid-state batteries (ASSBs) are widely believed to be promising for EVs because of their high safety and large volumetric energy density (Wh m⁻³) that are achieved using a non-flammable/non-volatile solid electrolyte (SE) and stacked design within a single external cell, respectively.^9–15 Among several types of ASSBs, oxide and sulfide types, in which oxide SE and sulfide SE are used, respectively, have been widely studied. The progress of the former in the application of EV batteries has been relatively slower than that of the latter, even though it offers superior chemical stability.^15–17 The major reason for this is the high interfacial resistance (R_int) between SE and the active material. Oxide SE is very hard and a sintering process at ∼1273 K is usually necessary to improve the connection among particles. However, such a high temperature sintering often promotes side reactions between active materials and SEs, resulting in the formation of impurity phases at the interfaces.

In contrast to oxide-type SEs, sulfide-type SEs have better ability to form interfaces with active materials because of their relatively soft nature, and they deform easily to form a densely packed structure with low R_int just by cold pressing. Thus far, many sulfide SEs whose ionic conductivity exceeds 1 mS cm⁻¹ have been reported.^18–21 Although this conductivity is lower than those of liquid electrolytes (∼10 mS cm⁻¹), a significant advantage for SEs is their transport number being unity, i.e., only Li ions contribute to the current flow. Considering that the transport numbers of many liquid electrolytes are around 0.3–0.5,²² sulfide-type ASSBs can work like conventional LIBs.

Studies on improving ionic conductivity are ongoing, and some SEs with ionic conductivities exceeding 10 mS cm⁻¹ have been reported by Kanno's group.^23,24 By utilizing these SEs, the power performance of ASSB may exceed that of LIBs largely in the future.

To increase the energy density beyond the current LIBs, anode-free type batteries, in which the Li metal negative electrode (NE) layer is electrochemically in situ formed onto the current collector during the initial charging process, have been extensively studied.^22,23 This system increases the energy density by lowering the weight of material components inside the batteries. Anode-free type is more favorable in combination with ASSBs (AF-ASSBs) because SE does not generate solid electrolyte interphase (SEI) and exhaustion of electrolyte by subsequent destruction and rebuilding of SEI according to the cycle does not occur. Unfortunately, most anode-free type batteries, including the AF-ASSB system, exhibit relatively low cycle capability because of the difficulty of the stable/reversible Li deposition/dissolution on the current collector during the charge–discharge process.^25–28 The largest difficulty is the inhomogeneous growth of the Li metal during the charge (also known as Li dendrite growth), which leads to low coulombic efficiency and sometimes short-circuit with penetrating the SE layer and reaching the positive electrode (PE). Although several approaches have been developed to reduce Li dendrite growth, this phenomenon has not been sufficiently suppressed.

Recently, however, a composite layer consisting of Ag nanoparticles and carbon black (CB) particles has been developed to improve the cycling capability of AF-ASSB. In a previous study, an AF-ASSB with a thin Ag/C layer between the SE layer and negative current collector was reported to have a high energy density (>900 W L⁻¹) and cycle capability (∼1000 cycles).^29,30 In this battery design, the deposited Li metal layer is formed between the Ag/C layer and the current collector. Suzuki et al.³⁰ explained that this phenomenon is caused by the overpotential of Li deposition. In the presence of this overpotential, Li deposition does not occur immediately even after the reaction potential drops below the Li potential (0 V vs. Li⁺/Li). The CB layer continuously accumulates Li atoms and becomes oversaturated. Li precipitates from the CB particles as a result of phase separation of the oversaturated CB (not formed directly from the reduction of Li⁺ in the SE layer), and this precipitated Li does not contribute to Li dendrite growth.

This explanation, however, has not been supported by experimental results; consequently, the details of this phenomenon remain unclear. To understand this further, we believe that investigating the reactions and associated structural changes in carbon (CB) during the charge–discharge processes is essential. Spencer-Jolly et al.³¹ performed an operando X-ray diffraction (XRD) measurement of the Ag/C (graphite) anode. They fabricated ASSBs with the Ag/C (graphite) anode, observed the structural evolution of Li–Ag and Li–C during charge and discharge at low and high current densities and usefully discussed the electrode reactions. In their report, however, they used graphite for the Ag/C anode for XRD, but the reported Ag/C anode consisted of CB, and it is desirable to observe a CB-based Ag/C anode. In this study, we used operando Raman spectroscopy, which allows for the observation of vibrational modes, bonding states, and structural changes by tracking Raman scattering processes during charge–discharge reactions^32–37 to investigate the role of the thin Ag/C layer in an AF-ASSB. In particular, the application of microscopic Raman spectroscopy to the cross section of AF-ASSBs is anticipated to identify slight morphological changes and metastable structural states on the Ag/C layer surface under dynamic conditions. Therefore, operando Raman spectroscopy was applied to the cross-section of an AF-ASSB during charge–discharge operations to clarify the electrode reactions of the Ag/C layer by evaluating the associated microscopic morphological and structural changes.

Experimental

Preparation of AF-ASSBs

The AF-ASSB system was fabricated, as described previously.³⁰ We used argyrodite-type lithium phosphorus sulfur chloride (Li₆PS₅Cl), whose ionic conductivity is ∼2 mS cm⁻¹ at room temperature for the SE.

A LiNi_0.9Co_0.05Mn_0.05O₂ was used for the PE active material known as a high-nickel NCM system. A Li₂O–ZrO₂ (LZO) thin layer was coated on the surface of the NCM particles using a sol–gel method, where the LZO sol was prepared from 2-propanol, lithium methoxide, and zirconium(IV) tetrapropoxide in a molar ratio of 200 : 2 : 1. In this study, we used a rolling fluidized coating machine (SFD-01, Powrex) to spray the sol onto the surface of the NCM particles. After coating, the sample was heated at 623 K for 1 h under air. The details of the coating are described elsewhere.^38,39

The PE powders thus obtained were mixed with SE powder, carbon nanofiber as a conductive agent, and a poly(tetrafluoroethylene) (PTFE) binder in a weight ratio 85 : 15 : 3 : 1.5. The resulting mixture was molded into sheet form, cut into a square with a length of about 1.7 cm (∼3 cm² area) and pressed onto an Al foil with 18 μm thickness as a current collector to form a composite PE. For the SE layer, the Li₆PS₅Cl powder was mixed with 1 wt% acrylic binder and stirred, while xylene and diethylbenzene were added to prepare the slurry. The slurry was coated onto a non-woven fabric using a blade coater and dried in air at 313 K. The thickness of the SE sheet was ∼100 μm. The resulting stacked structure was dried in a vacuum state at 313 K for 12 h and cut into a square with a length of about 2.2 cm. The preparation of the PE and SE sheets was performed in a dry room (dew point: ∼223 K).

The Ag/C layer was fabricated in the following manner. First, we placed 12 g of CB (D₅₀ = 38 nm) and 4 g of Ag (D₅₀ = 60 nm) in a container with 14 g of an N-methyl-pyrrolidone (NMP) solution, including 8 wt% of a polyvinylidene fluoride. Next, the mixture was stirred while slowly adding 50 g of NMP to prepare an Ag/C slurry. The obtained slurry was coated onto a stainless steel (SUS) foil of 10 μm thickness using a blade coater, dried in air at 353 K for 20 min, and subsequently dried under vacuum at 373 K for 12 h in a dry room. They were cut into a square with a length of about 2 cm. The loading was adjusted so that the capacity of the Ag/C layer is ∼10% that of the PE layer. The PE, SE and Ag/C layers were stacked together in this order and encapsulated in a laminating film in a vacuum state to fabricate an ASSB. This stacking was done in the dry room. The ASSB was subjected to warm isostatic pressing at 490 MPa and 353 K for 30 min. The capacity of the prepared AF-ASSB was designed at approximately 18 mA h at a C/10 rate. The thicknesses of the PE, SE and Ag/C layers were approximately 70, 100 and 10 μm, respectively.

Operando Raman spectroscopy

Operando Raman spectroscopy measurements for the Ag/C layer were carried out under potential sweep conditions with a scan rate of 0.11 mV s⁻¹, within an open-circuit voltage range from 4.0 to 2.1 V at room temperature. The laminate-type AF-ASSB with Ag/C layer was cut and introduced into the electrochemical measurement cell with an observation window of quartz glass under a suitable pressure (∼2 MPa) in an Ar-filled glove box (dew point: ∼185 K). The electrochemical measurement cell containing AF-ASSB was connected to a potentiostat (HZ-7000, Meiden Hokuto) inside a microscope-type Raman spectrometer (NRS-4500, Jasco). Raman spectroscopy was performed using an excitation wavelength of 532.0 nm, laser power of 3.2 mW, 50× objective, pinhole slit of ∅ 34 μm, and grating of 900 mm⁻¹ (laser size: ∼1 μm, optical resolution: ca. 4.7 cm⁻¹). All Raman spectra were acquired with 60 s of irradiation and four accumulations. Raman spectra were acquired from two points inside the Ag/C layer to ensure the reliability of the data, as shown in Fig. 1. The obtained Raman spectra were corrected using the peak positions of polypropylene as a standard sample.

Fig. 1 Appearance of the AF-ASSB of a laminate-type cell (a) and microscopic image of its cross-section (b).

Prior to the operando Raman spectroscopy measurements, the constant current-constant voltage (CCCV) charge–discharge test was also carried out using an AF-ASSB fabricated using the same method as the above cell to evaluate the typical electrochemical properties. The test was performed with a current of 5 mA, voltage range of 4.2–2.5 V, and CV of 4.2 V for 2 h at 318 K using a restraint cell under approximately 2 MPa.

Ex situ Raman spectroscopy

To compare with operando Raman spectroscopy, ex situ Raman spectroscopy was also performed on the AF-ASSB with an Ag/C layer. In ex situ Raman spectroscopy, each Raman spectrum was acquired from the cross-section of the cell after charging up to a determined state-of-charge (SOC). First, the cell was cut and introduced into the same cell for operando Raman spectroscopy in an Ar-filled glove box. The cut AF-ASSB was subjected to constant-current charging or discharging at 0.5 mA for 4 h (approximately 2 mA h) at 318 K and then maintained in an open-circuit condition for 5 h to allow for the relaxation of the electrochemical reactions. Raman spectroscopy was performed on each charged or discharged state after maintaining the open-circuit conditions. Raman spectra were acquired from the Ag/C layer at 10 points across the width direction (approximately every 2 mm) using the same measurement conditions as operando spectroscopy at room temperature.

Results and discussion

Electrochemical and spectroscopic properties of the AF-ASSB

Typical electrochemical properties were examined using a constant current-constant voltage charge–discharge test at a C/4 rate (5 mA) at 318 K, as shown in Fig. 2(a) and (b). The charge–discharge profile exhibited a sufficient discharge capacity of about 16 mA h at the 1st cycle despite the presence of irreversible capacities between the charge and discharge processes. Additionally, the discharge capacities monotonically decreased with cycle number owing to the higher operating current compared to a previous study that examined a C/10 rate.²⁹ The coulombic efficiencies were approximately 81% and 95% in the 1st and 2nd cycles, respectively, and then remained over 99.98% for up to 300 cycles.

Fig. 2 Typical charge–discharge profile (a) and cycle properties (b) of the AF-ASSB with the Ag/C layer and Raman spectra of the SE, CB and cross-sectional Ag/C layer in a pristine cell (c).

Fig. 2(c) shows the Raman spectra for the CB powder, which is the same powder used for the Ag/C layer, and the SE and the Ag/C layers in a pristine state in the cross-section AF-ASSB under the Ar atmosphere. Raman spectra in the CB powder and Ag/C layer in a pristine state were separately acquired using the sealed and electrochemical measurement cells, respectively. In the SE layer, an intense peak at 426 cm⁻¹ was observed, which was derived from the PS₄³⁻ species in the Li₆PS₅Cl argyrodite.^40,41 The CB powder displayed D- and G-bands at 1349 and 1589 cm⁻¹, respectively. These D-(A_1g) and G-(E_2g) bands agree with the characteristics of the disordered carbon materials because of their wide peak widths and similar intensities. Such broad peaks of the D-(A_1g) and G-(E_2g) bands are known to be characteristic of disordered carbon with partially ordered graphene layers (low crystallinity).^41–44

Raman spectra of the Ag/C layer in the cross section (upper in Fig. 2(c)) clearly displayed D- and G-bands at the same positions as those of the CB powder. Therefore, we consider that no structural change or disintegration has occurred during the cell fabrication process and the preparation process for the measurement, such as cutting the AF-ASSB and introducing it into the electrochemical measurement cell.

Operando Raman spectroscopy for the AF-ASSB with Ag/C layer

The electrochemical profiles obtained for the potential sweep during operando Raman spectroscopy at room temperature are illustrated in Fig. S1.† The electrochemical profiles showed a relatively flat current response below 3.0 V although a small peak appeared around 0.7 V during the charging process (Fig. S1(b)†). The main current peak around 3.7 V was ascribed to the deintercalation of Li⁺ from the active material (NCM) in the PE with respect to the deposition of metallic Li on the SUS current collector, suggesting that a redox reaction certainly occurred during the charge. The spike current response of over 15 mA observed at around 3.9 V was attributed to the short-circuit of the cell. Notably, the short-circuit was also confirmed in other operando Raman spectroscopy measurements (n = 3), indicating that the short-circuit most likely occurred through the cross-sectional surface rather than inside the cell. The short-circuited cell, however, possessed a stable open-circuit voltage of around 3.7 V and was continuously discharged to 2.1 V with a clear current peak of around 3.6 V, as depicted in Fig. S1.† Therefore, we consider that the effect of the short-circuit was considerably small and that the charge–discharge reaction proceeded even after the short-circuit, justifying the operando Raman measurement for this cell.

Fig. 3 shows the Raman spectra measured at point 1 (see Fig. 1) in the cross-section Ag/C layer of the AF-ASSB at each voltage via operando Raman spectroscopy (the results for point 2 are also shown in Fig. S2 in the ESI†). Note that all Raman spectra in Fig. 3 are presented as raw data without baseline correction to clearly show the effect of fluorescence. Additionally, the trends of Raman spectra at points 1 and 2 have similar features, such as the fluorescence effect, change in the D- and G-bands and additional peaks, discussed in the subsequent section. Therefore, the dependency on the measurement positions in the obtained Raman spectra is expected to be minimal. The changes in cell volume were estimated to be approximately ±5 μm at the cross-section by the optical images during the charge–discharge process, and measurement points were also maintained by the adjustment of the x–y stage.

Fig. 3 Raman spectra acquired from point 1 in the cross-sectional AF-ASSB with the Ag/C layer at every 0.3 V interval during charge–discharge processes in the operando Raman spectroscopy at room temperature.

At both measuring points, the Raman spectra displayed an obvious fluorescence effect that caused the disordered baseline during charging (low wavenumber region, as illustrated in Fig. 3 and S3†). The fluorescence effect usually depends on the Raman scattering processes associated with changes in electron states, polarizability, or excitation level induced by laser irradiation.^45,46 This effect is often observed in metal, alloy, and impurity components. Therefore, the observed Raman spectra with fluorescence likely reflect the morphological change in the carbon material to a metallic-like state in the Ag/C layer caused by metallic Li deposition. Furthermore, the shape of the D- and G-bands at 1347 and 1583 cm⁻¹, respectively, changed compared to that of a pristine state, especially in the range of about 3.3 V (charging) to 2.7 V (discharging) at both measurement points. In addition to this change, a new band appeared between the D- and G-bands in the Raman spectra in this SOC region, as shown in Fig. S4.†

As a quantitative analysis of the peak shape of the D- and G-bands, curve fitting based on the pseudo-Voigt function was carried out for the Raman spectra in the range of 700–1850 cm⁻¹ (Fig. S5†). Notably, the curve fitting was performed for only two components (D- and G-bands) because of their complex peaks and relatively low signal-to-noise ratio. Furthermore, this region may include sub-peaks, such as D2-, D3- and D4-bands; however, the obtained Raman spectra did not clearly show these components (see explanation in ESI†).⁴⁷ The exact calculation of the peak intensities was difficult because of the lack of peaks for y-axis correction and normalization. The calculated parameters are shown in Tables S1 and S2.† These values varied within a relatively wide range possibly owing to factors such as signal-to-noise ratio, hidden sub-peaks, and other spectral components. However, the formation of a new band in the D- and G-band regions was also suggested owing to changes in peak shape, as depicted in Fig. S3,† and the aspect ratio discussed below. To analyze the peak shape changes, the sum of the peak heights of the D- and G-bands divided by the sum of peak widths of the D- and G-bands was calculated as the ‘aspect ratio’ in this study, as described in Fig. S6.† The calculated aspect ratios at each voltage in points 1 and 2 are shown in Fig. 4(a). The maximum aspect ratios were observed at around 2.1–2.4 V at both measurement points during the charge–discharge processes. This behavior should support the change in the peak shapes of the D- and G-bands in the middle of the SOC region in the quantitative analysis based on curve fitting during charge–discharge reactions. The peak height ratio of the D- to the G-bands using the curve fitted parameters is commonly defined as R, which reflects the degree of structural regularity in carbon materials (see Fig. S6†).^43,44 Herein, R was calculated at each voltage for each measurement point, as shown in Fig. 4(b). Additionally, the details of the R value in carbon materials are discussed in the ESI.† Although the trends of R were different for each measurement point in the charge process, these values decreased during the discharge process in all voltage ranges. These trends probably reflect the structural and regularity changes in the graphene layer inside the CB caused by the deposition of metallic Li during the charge process. We also calculated the changes in the peak positions of the D- and G-bands from the curve fitting analysis, as shown in Fig. 4(c) and (d). At both measurement points, the position of the D-band, especially at point 2, increased up to approximately 1.5 V, then slightly decreased up to approximately 2.4 V, and finally increased further during the charge process, while the position increased continuously during the discharge process. The position of the G-band slightly decreased continuously below approximately 3.0 V and then decreased considerably at a higher voltage during the charge process, whereas a substantial increase in all voltage regions was observed during the discharge process. The gradual changes (especially for the G-band position) should correspond to the partial intercalation/deintercalation reaction of Li⁺. There is a possibility that both measurement points changed continuously during the charge–discharge process because of the volume changes associated with the deposition and dissolution of metallic Li. However, the changes in aspect ratios, Rs, and positions of the D- and G-bands suggest that the structure/regularity of CB (e.g., bond angle/length, bond disorder, and clustering order) was affected by electrochemical reactions, such as the small amount of intercalation/deintercalation and adsorption/desorption processes regardless of the disordered carbon material. A similar peak shift has been reported for a graphite negative electrode active material, which was interpreted as the formation of a stage structure of the Li-graphite intercalated compounds (Li-GICs).^44,48–51 In several studies, the Li-GIC structure in the liquid electrolyte has been investigated, and the phase transition from stage 4 to stage 1 in the high SOC region (low potential region vs. Li metal) has been explained by operando Raman spectroscopies. Here, the stage structure is often described as changing the number of graphene layers that separate Li⁺ during the intercalation reaction. For instance, in stage 4, Li⁺ is separated by 4 graphene layers (LiC₃₆). By transitioning to stage 1, the G-band begins to shift to a lower wavenumber and split of peak shapes, and then to a broader shape, associated with the twist of the graphene layer and strain by Li⁺ intercalation. In this study, as similar trends were observed, CB was also expected to intercalate Li and change the structural regularity in graphene layers, resulting in the changes in the R values and peak positions in the D- and G-bands.

Fig. 4 Relationships between the aspect ratios (a), R values (b), and positions (c and d) in the D- and G-bands on voltage calculated using the curve fitting analysis.

In addition to the fluorescence effect and changes in the D- and G-bands in the Raman spectra, three additional peaks approximately 1087, 1854, and 3648 cm⁻¹ (green regions in Fig. 3) appeared during charging, and that at 1854 cm⁻¹ also remained after the discharging. In a previous report, transmission electron microscope and electron diffraction pattern observations indicated that an Ag–Li alloying reaction with charging occurred in the Ag/C layer of an AF-ASSB system with the same cell configuration.²⁷ Therefore, the Ag–Li alloying reaction also occurred in this AF-ASSB. This charging mechanism may affect the obtained Raman spectra, such as the observed additional peaks.

Microscopic images and Raman spectra of the cross section of the AF-ASSB were acquired after short-circuit (∼3.9 V) and in the discharged state. Fig. 5(a) shows cross-sectional images of the AF-ASSB in the pristine state, at 3.9 V (after short-circuit), and at 2.1 V (discharged) and Raman spectra of the AF-ASSB in these states obtained by operando Raman spectroscopy. The cross-sectional image of the SE layer clearly exhibited line-like cracks and small precipitates with metallic luster close to the Ag/C layer after the short-circuit; these features mostly remained after the discharge process. Moreover, the microscopic image revealed that the PE layer had a slight metallic luster after a short-circuit compared to that of the pristine state. We consider that these cracks and precipitates induced the short-circuit in the case with the cross-section because the laminate-type cell before the cut exhibited stable charge–discharge behavior and cycle capability (Fig. 2(b)). The Raman spectra acquired from the precipitate area also exhibited a peak of around 1848 cm⁻¹ (the green region in Fig. 5(b)), which was not observed in a pristine state. Since this peak appears at a wavenumber similar to the new peak that formed during the charge process (Fig. 3), it is attributed to some Raman active species induced from the precipitate with a metallic luster. In general, metals and alloys hardly exhibit Raman scattering owing to the dominance of Rayleigh scattering and high background signals. However, since the formed precipitates were predicted to be small similar to the laser spot size, the obtained Raman spectra were expected to include both information on the surface and the neighboring area of the Ag–Li particles. In this experiment, the additional and precipitate peaks may correspond to compounds related to Ag, CB, and Li. In addition, this formed compound could diffuse physically into the SE layer owing to insufficient pressure at the cross-section. Note that this peak at 1848 cm⁻¹ can hardly be attributed to some Li₆PS₅Cl decomposed products because the measurement point has a metallic luster. Furthermore, in a previous report, the reductive byproduct of Li₆PS₅Cl exhibits a peak at 383 cm⁻¹, corresponding to Li₂S, and is separated from the peak at 1848 cm⁻¹ and D- and G-bands.⁵² Therefore, although these additional peaks are associated with Ag–Li alloying reactions, the precise assignment of these additional peaks remains difficult.

Fig. 5 Microscopic images of the cross-sectional AF-ASSB at pristine, 3.9 V (charge) and 2.1 V (discharge) states during operando Raman spectroscopy (a), and their Raman spectra (b) acquired from the observed precipitates in the SE layer after the short-circuit and at points 1 and 2 during potential sweep.

Comparison with ex situ Raman spectroscopy

To compare with the operando Raman spectroscopy, ex situ Raman spectroscopy was performed for the cross-section of the AF-ASSB with the Ag/C layer. Although the obtained capacity was insufficient compared to that of the without cut cell (Fig. 2(a)), the charge–discharge profile indicates the electrochemical reactions with metallic Li deposition/dissolution, with nonlinear behaviors (constant-current period), as shown in Fig. S7.† Additionally, the charge and discharge capacities during the constant-current period were confirmed to be approximately 10 mA h, corresponding to 56% of the theoretical value. The deposited metallic Li layer was clearly observed between the Ag/C layer and the SUS current collector in the cross-sectional view of the AF-ASSB at every SOC (Fig. S8†). Here, Li⁺ may deposit onto the SUS current collector because it has more space there than between the SE and Ag/C layers. In addition, voltages in the open-circuit period were maintained constant for 5 h and reached sufficient relaxing states. Fig. 6 shows the Raman spectra at every SOC acquired from the Ag/C layer after being maintained under the open-circuit condition (represented data in the 10 points are shown). Similar to operando Raman spectroscopy, the fluorescence effect in the lower wavenumber region was observed with higher SOC, suggesting a change in the metallic-like state with Li accumulation in the Ag/C layer. In contrast to the operando measurement, an additional peak was observed only at 3650 cm⁻¹ in the charge process and then disappeared in the reversible discharge process (the green region in Fig. 6).

Fig. 6 Raman spectra acquired from the cross-sectional AF-ASSB with the Ag/C layer of every SOC during charge–discharge processes in the ex situ Raman spectroscopy.

However, as we mentioned in the previous section, we observed three activated species as additional peaks at 1087, 1854, and 3648 cm⁻¹ (Fig. 3) with the operando Raman measurement, but we did not observe the former 2 peaks with the ex situ measurement. This suggests that the formation process of the activated species for Raman scattering with metallic Li deposition for the ex situ measurement was different from that in the operando Raman spectroscopy with continuous charging. One of the reasons for this phenomenon is that the additional peaks observed in the operando measurement correspond to the metastable state of the Ag/C layer during charging. Since this ex situ measurement was performed after relaxation, these peaks may not be observed under steady-state conditions.

To analyze the Raman spectra quantitatively, curve fitting was performed using the pseudo-Voigt function, following the same method as in the operando measurements. In this analysis, three spectra with relatively low fluorescence were selected from the ten measured points, and the fitted parameters were averaged to reduce the influence of large background fluorescence. The calculated values are shown in Tables S3 and S4.†Fig. 7 shows the calculated aspect ratio (a), R value (b) and position (c) of D- and G-bands as average values of 3 measurement points in the ex situ Raman spectroscopy. The aspect ratios exhibited a slightly continuous increasing trend in the charge process and then decreased in the discharge process although their variation values were lower than those of data calculated from the operando measurement (Fig. 4(a)). Furthermore, the R values confirmed the continuous decrease/increase trend in the charge and discharge processes, respectively. Although the positions of the D- and G-bands did not show clear trends, the variation from the initial state was also observed by the SOC changing. These results suggest that the morphological and structural changes in the Ag/C layer are slight compared with the operando measurement, even though metallic Li deposition occurs with the charge. In this ex situ measurement, the charge–discharge operation was performed via a constant-current charge–discharge test that was maintained under an open-circuit condition (relaxed state) and was expected to reach a sufficient steady-state at every SOC. In contrast, the operando measurement under dynamic conditions continued the electrochemical reactions via the potential sweep operation and insufficiently reached a steady-state for every SOC. This unsteady-state in operando Raman spectroscopy should affect the morphological and structural changes in the Ag/C layer, such as the different additional peaks and clear changes in the D- and G-bands, compared with ex situ measurement. In particular, this operando measurement was performed under a relatively high scan rate of potential sweep and room temperature and may have partially formed an oversaturated state in the Ag/C layer concentrated Li⁺ with charging. Therefore, ex situ Raman spectroscopy basically supports the results of the operando measurement, and different behaviors can be explained by the kinetic changes, such as reaching a steady-state or not. In addition, operando Raman spectroscopy can help to reveal the unstable state formed temporarily by charge–discharge operations, represented as additional peaks and clear changes in the carbon structures with charging, because of their dynamic condition.

Fig. 7 Relationships between aspect ratios (a), R values (b), and positions (c) in the D- and G-bands calculated using the curve fitting analysis. All data are described as average values of the 3 measurement points.

From the results of operando/ex situ Raman spectroscopy, the electrode reaction with the Ag/C layer shown in Fig. 8 was proposed. When the Li–Ag alloy is formed, the charge reaction in the AF-ASSB is estimated to occur in the following steps.

Fig. 8 Schematic of the reaction model in the cross-sectional AF-ASSB with an Ag/C layer during charging process.

(1) In the low SOC region, Li⁺ migrates from the PE to Ag/C layers and precipitates as a metallic Li layer between the Ag/C layer and the SUS current collector with Li–Ag alloying. When Li⁺ migrates into the Ag/C layer, structural changes in CB also occur, such as intercalation and adsorption reactions for the inside/surface of the CB particles.

(2) In the middle SOC region, the thickness of the metallic Li layer expands as the charge reaction proceeds.

(3) In the high SOC region, the growth of the metallic Li layer with consecutive phase transitions of Li–Ag alloys continued. Although short-circuits often occur in the case of the cross-section, the Ag/C layer continues the stable Li deposition by the Ag–Li alloying and interaction between the CB and Li⁺ and is expected to suppress Li dendrite formation, even with though higher Li⁺ concentration compared with the low SOC region.

The above results suggest that the short-circuit occurs in the vicinity of the cross-section of the AF-ASSB. However, the reaction of CB with Li⁺ and Ag–Li alloying can also play a role in relaxing the current concentration with the Li deposition during the initial charging because of the amount of electricity used for these processes to proceed. These functions are expected to contribute to stable charge–discharge and cycle capability for the long term in AF-ASSB. Therefore, operando Raman spectroscopy can reveal continuous structural changes in the disordered carbon material with partial Li⁺ intercalation/deintercalation in the Ag/C layer of AF-ASSBs. Moreover, kinetic differences for the structural changes in the disordered carbon materials depending on the degree of steady-state could be discussed by comparing both operando/ex situ Raman spectroscopies.

In the next step, we investigate to assign for the additional peaks in the operando Raman spectra precisely using some model cells, such as a Li symmetrical cell, to give a complete explanation of the dynamical reaction of Li in the Ag/C layer to elucidate the role of Ag/C layer and charge–discharge mechanism in an AF-ASSB.

Conclusions

The structural changes in the Ag/C layer during charge–discharge were investigated using operando/ex situ Raman spectroscopies. In operando Raman spectroscopy under potential sweeping, the Raman spectra exhibited a fluorescence effect in the lower wavenumber region during charging, indicating a transition to the metallic state in the Ag/C layer with Li deposition. The D- and G-bands related to CB clearly showed clear changes from the initial state during the charge–discharge process. From the curve fitting analysis for the D- and G-bands, both R values and positions exhibited significant changes at middle to high voltage. This trend was similar to that reported in graphite materials, suggesting Li⁺ intercalation into CB despite its amorphous (disordered) structure. Furthermore, Raman spectra with charging showed three additional peaks associated with the Ag–Li alloying reaction. In contrast, ex situ Raman spectroscopy under a relaxed state exhibited a gradual change in the D- and G-bands and only a subset of the additional peaks compared to the operando measurement. This difference should correspond to dynamic (operando) or static (ex situ) states in the Ag/C layer owing to achieving a non-equilibrium reaction under potential sweeping. Therefore, we demonstrated that the CB undergoes structural changes with Li⁺ intercalation despite the disordered carbon material, and these changes occur more rapidly under dynamic conditions than under static conditions.

Data availability

The data supporting this article have been included as part of the ESI.†

Author contributions

K. H. performed the operando Raman spectroscopy and electrochemical measurements and prepared the draft of the manuscript. N. S. and J. S. fabricated the AF-ASSB cells and supported the evaluations. S. S. supervised this research. All authors contributed to the discussion of the results and commented on the manuscript.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

We acknowledge Ms Yuna Sato at Kogakuin University for supporting the operando/ex situ Raman spectroscopies and electrochemical measurements.

Notes and references

1. B. Obama, Science, 2017, 355, 126–129.
2. B. Dunn, H. Kamath and J.-M. Tarascon, Science, 2011, 334, 928–935.
3. V. Etacheri, R. Marom, R. Elazari, G. Salitra and D. Aurbach, Energy Environ. Sci., 2011, 4, 3243–3262.
4. A. Yoshino, Angew. Chem., Int. Ed., 2012, 51, 5798–5800.
5. M. Winter, B. Barnett and K. Xu, Chem. Rev., 2018, 118, 11433–11456.
6. C. Xu, P. Behrens, P. Gasper, K. Smith, M. Hu, A. Tukker and B. Steubing, Nat. Commun., 2023, 14, 119.
7. Y. Tian, G. Zeng, A. Rutt, T. Shi, H. Kim, J. Wang, J. Koettgen, Y. Sun, B. Ouyang, T. Chen, Z. Lun, Z. Rong, K. Persson and G. Ceder, Chem. Rev., 2021, 121, 1623–1669.
8. H. Zhang, X. Hu, Z. Hu and S. J. Moura, Nat. Rev. Electr. Eng., 2024, 1, 35–52.
9. W. Hou, X. Guo, X. Shen, K. Amine, H. Yu and J. Lu, Nano Energy, 2018, 52, 279–291.
10. Y. Gambe, Y. Sun and I. Honma, Sci. Rep., 2015, 5, 8869.
11. T. Liu, Y. Yuan, X. Tao, Z. Lin and J. Lu, Adv. Sci., 2020, 7, 2001207.
12. L. Fan, S. Wei, S. Li, Q. Li and Y. Lu, Adv. Energy Mater., 2018, 8, 1702657.
13. T. Famprikis, P. Canepa, J. A. Dawson, M. S. Islam and C. Masquelier, Nat. Mater., 2019, 18, 1278–1291.
14. J. Janek and W. G. Zeier, Nat. Energy, 2023, 8, 230–240.
15. A. Manthiram, X. Yu and S. Wang, Nat. Rev. Mater., 2017, 2, 16103.
16. V. Thangadurai, S. Narayanan and D. Pinzaru, Chem. Soc. Rev., 2014, 43, 4714.
17. Y. Xiao, Y. Wang, S. H. Bo, J. C. Kim, L. J. Miara and G. Ceder, Nat. Rev. Mater., 2021, 6, 868.
18. H. Tsukasaki, H. Morimoto and S. Mori, Solid State Ionics, 2020, 347, 115267.
19. F. Han, Y. Zhu, X. He, Y. Mo and C. Wang, Adv. Energy Mater., 2016, 6, 1501590.
20. H. Muramatsu, A. Hayashi, T. Ohtomo, S. Hama and M. Tatsumisago, Solid State Ionics, 2011, 182, 116–119.
21. Y. Liu, X. Xu, M. Sadd, O. O. Kapitanova, V. A. Krivchenko, J. Ban, J. Wang, X. Jiao, Z. Song, J. Song, S. Xiong and A. Matic, Adv. Sci., 2021, 8, 1–11.
22. K. Xu, Chem. Rev., 2004, 104, 4303.
23. N. Kamaya, K. Homma, Y. Yamakawa, M. Hirayama, R. Kanno, M. Yonemura, T. Kamiyama, Y. Kato, S. Hama, K. Kawamoto and A. Mitsui, Nat. Mater., 2011, 10, 682–686.
24. Y. Li, S. Song, H. Kim, K. Nomoto, H. Kim, X. Sun, S. Hori, K. Suzuki, N. Matsui, M. Hirayama, T. Mizoguchi, T. Saito, T. Kamiyama and R. Kanno, Science, 2023, 381, 50.
25. A. Pei, G. Zheng, F. Shi, Y. Li and Y. Cui, Nano Lett., 2017, 17, 1132–1139.
26. J. Xie, J. Wang, H. R. Lee, K. Yan, Y. Li, F. Shi, W. Huang, A. Pei, G. Chen, R. Subbaraman, J. Christensen and Y. Cui, Sci. Adv., 2018, 4, 2–9.
27. C.-M. Park, H. Jung and H.-J. Sohn, Electrochem. Solid-State Lett., 2009, 12, A171.
28. W. Liu, Y. Xia, W. Wang, Y. Wang, J. Jin, Y. Chen, E. Paek and D. Mitlin, Adv. Energy Mater., 2019, 9, 1802918.
29. Y. G. Lee, S. Fujiki, C. Jung, N. Suzuki, N. Yashiro, R. Omoda, D. S. Ko, T. Shiratsuchi, T. Sugimoto, S. Ryu, J. H. Ku, T. Watanabe, Y. Park, Y. Aihara, D. Im and I. T. Han, Nat. Energy, 2020, 5, 299–308.
30. N. Suzuki, N. Yashiro, S. Fujiki, R. Omoda, T. Shiratsuchi, T. Watanabe and Y. Aihara, Adv. Energy Sustain. Res., 2021, 2, 2100066.
31. D. Spencer-Jolly, V. Agarwal, C. Doerrer, B. Hu, S. Zhang, D. L. R. Melvin, H. Gao, X. Gao, P. Adamson, O. V. Magdysyuk, P. S. Grant, R. A. House and P. G. Bruce, Joule, 2023, 7, 503–514.
32. C. P. Grey and J. M. Tarascon, Nat. Mater., 2016, 16, 45–56.
33. Improving the fundamental understanding of batteries via operando measurements, Nature, 2022, 13, 4723,  DOI:10.1038/s41467-022-32245-9.
34. A. Jena, Z. Tong, B. Bazri, K. Iputera, H. Chang, S. F. Hu and R. S. Liu, J. Phys. Chem. C, 2021, 125, 16921–16937.
35. D. Liu, Z. Shadike, R. Lin, K. Qian, H. Li, K. Li, S. Wang, Q. Yu, M. Liu, S. Ganapathy, X. Qin, Q. Yang, M. Wagemaker, F. Kang, X. Yang and B. Li, Adv. Mater., 2019, 31, 1806620.
36. D. Cao, Y. Zhang, A. M. Nolan, X. Sun, C. Liu, J. Sheng, Y. Mo, Y. Wang and H. Zhu, Nano Lett., 2020, 20, 1483–1490.
37. J. Zhang, C. Zheng, L. Li, Y. Xia, H. Huang, Y. Gan, C. Liang, X. He, X. Tao and W. Zhang, Adv. Energy Mater., 2020, 10, 1903311.
38. S. Ito, S. Fujiki, T. Yamada, Y. Aihara, Y. Park, T. Y. Kim, S.-W. Baek, J.-M. Lee, S. Doo and N. Machida, J. Power Sources, 2014, 248, 943.
39. N. Suzuki, W. D. Richards, Y. Wang, L. J. Miara, J. C. Kim, I.-S. Jung, T. Tsujimura and G. Ceder, Chem. Mater., 2018, 30, 2236.
40. T. Hakari, M. Deguchi, K. Mitsuhara, T. Ohta, K. Saito, Y. Orikasa, Y. Uchimoto, Y. Kowada, A. Hayashi and M. Tatsumisago, Chem. Mater., 2017, 29, 4768–4774.
41. P. A. Marsh, A. Voet, T. J. Mullens and L. D. Price, Carbon, 1971, 9, 797–805.
42. M. Seredycha, D. H-Jurcakovab, G. Q. Lub and T. J. Bandosza, Carbon, 2008, 46, 1475–1488.
43. A. C. Ferrari and J. Robertson, Phys. Rev. B:Condens. Matter Mater. Phys., 2000, 61, 14095–14107.
44. A. C. Ferrari, Solid State Commun., 2007, 143, 47–57.
45. P. Johansson, H. Xu and M. Kall, Phys. Rev. B:Condens. Matter Mater. Phys., 2005, 72, 035427.
46. S. Hernandez, M. Perez-Estebanez, W. Cheuquepan, J. V. Perales-Rondon, A. Heras and A. Colina, Anal. Chem., 2023, 95, 16070–16078.
47. V. Sharma, D. Uy, A. Gangopadhyay, A. Neill, W. A. Paxton, A. Sammut, M. A. Ford and P. B. Aswath, Carbon, 2016, 103, 327–338.
48. J. C. Chacón-Torres, L. Wirtz and T. Pichler, ACS Nano, 2013, 7, 9249–9259.
49. S. Maruyama, T. Fukutsuka, K. Miyazaki and T. Abe, Electrochim. Acta, 2018, 265, 41–46.
50. C. Sole, N. E. Drewett and L. J. Hardwick, Faraday Discuss., 2014, 172, 223–237.
51. H. Nakagawa, Y. Domi, T. Doi, M. Ochida, S. Tsubouchi, T. Yamanaka, T. Abe and Z. Ogumi, J. Power Sources, 2012, 206, 320–324.
52. Y. Zhou, C. Doerrer, J. Kasemchainan, P. G. Bruce, M. Pasta and L. J. Hardwick, Batter. Supercaps, 2020, 3, 647–652.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5ta02406d

---