Syed Atif
Pervez‡
*a,
Bhaghavathi P.
Vinayan‡
*a,
Musa Ali
Cambaz
a,
Georgian
Melinte
b,
Thomas
Diemant
c,
Tobias
Braun
a,
Guruprakash
Karkera
a,
R. Jürgen
Behm
ac and
M.
Fichtner
*ab
aHelmholtz Institute Ulm, Helmholtzstraße, 11, 89081, Ulm, Germany. E-mail: syedatif.pervez@partner.kit.edu; vinayan.parambath@kit.edu; m.fichtner@kit.edu
bInstitute of Nanotechnology (INT), Karlsruhe Institute of Technology (KIT), Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany
cInstitute of Surface Chemistry and Catalysis, Ulm University, Albert-Einstein-Allee 47, 89081 Ulm, Germany
First published on 14th July 2020
The Li+ transport kinetics at the solid–solid electrode|electrolyte interfaces are crucial for the stable and durable performance of solid-state batteries (SSBs). A poor interface due to mechanical problems and/or (electro-)chemical instabilities will curtail the performance of such batteries. Herein, we present a detailed study on the interfaces of a lithium–sulfur (Li–S) SSB with a Li anode, Li–garnet (LLZO) solid electrolyte (SE), and a sulfur–carbon composite as the cathode. Interlayer gels based on ionic liquids were introduced to improve the interfacial properties of the system. For Li symmetric cells, the strategy resulted in a decrease in cell resistance by about a factor of five and stable voltage profiles with low overpotentials (∼300 mV at 0.4 mA cm−2 after >450 hours). Furthermore, the LLZO SE efficiently blocked the polysulfide shuttle to the Li anode. Due to the advantageous features of the design, good electrochemical performance was obtained, where the Li–S SSB delivered an initial discharge capacity of ca. 1360 mA h gsulfur−1 and a discharge capacity of ca. 570 mA h gsulfur−1 after 100 cycles. Detailed electrochemical and compositional characterization of the interphase layers was performed at the Li anode and sulfur cathode interfaces through X-ray photoelectron spectroscopy (XPS), applying depth-profiling techniques, and scanning transmission electron microscopy (STEM). The results revealed the presence of interphase nano-layers with varying thicknesses on the LLZO surface which contained organic and inorganic species.
Utilization of ceramic SE based separators is also an interesting strategy that can potentially solve problems associated with the S cathode and Li anode. Thanks to their high shear modulus, both the polysulfide shuttle and the Li dendrite growth can be mechanically suppressed. Among various inorganic ceramic SEs, Li–garnet (LLZO) is of particular interest since it offers a relatively high (electro)-chemical stability with Li metal, a much wider potential window compared to other SEs,11 and a high shear modulus (61 GPa for LLZO; 4.2 GPa for Li).12 However, the surface impurities/roughness and brittle nature of LLZO SE restrict its intimate contact with Li and other solid-state cathode materials.13,14 With S as the cathode material, the electrode/electrolyte integration will be even more challenging considering the enormous volumetric expansion during the conversion reaction with Li. In this regard, surface modification strategies including the use of thin nanometric layer coatings, and 3D interfacial architectures, or the implementation of polymer, gel and liquid interlayers have been used to keep the electrode|SE contacts intact and stabilize the interfacial ionic transport.15–18 Using Li ion conducting gel/liquid interlayers instead of using solid interlayers may prove to be an even better strategy due to their inherent tendency to penetrate through the porous structures and hence to access a greater surface area of the composite cathode.19
Herein, we have designed a solid-state Li–S storage system based on a sulfur–carbon composite positive electrode, LLZO SE and a Li negative electrode. To improve the interfacial properties of the SSB, interlayer gels containing ionic liquids (IL) and a Li-salt have been introduced at electrodes|SE interfaces. The LLZO SE acts as a solid barrier to mechanically block the polysulfide shuttle. Owing to such features, good battery performance has been demonstrated, where the cells exhibited an initial discharge capacity of ca. 1360 mA h gsulfur−1 and a discharge capacity of ca. 570 mA h gsulfur−1 after 100 cycles. The chemical composition of the interphase layer, both of its surface and of deeper regions, was characterized by XPS, applying depth profiling techniques, including both sides (cathode & anode) of the LLZO SE. High-resolution transmission electron microscopy (HRTEM) and energy dispersive X-ray spectroscopy (EDS) spectrum imaging were also conducted to obtain information about the thickness and composition of the interphase layer. The data suggest the presence of amorphous interfacial nano-layers on the LLZO surface with varying thickness. The layers contain both organic and in-organic species originated from the decomposition of the IL, Li-salt and organic solvents.
(1) |
To identify the potential and role of the interlayers in improving the Li|LLZO interfacial properties, Li symmetrical cells (Li|LLZO|Li) were assembled with and without gel interlayers and EIS spectra were recorded at room temperature. As described in more detail in the experimental section, the interlayers were composed of the ionic liquid Pyr14TFSI (IL) with LiTFSI as Li salt and organic solvents. Special care was taken to use a minimal amount (5 μL) of gel ion conductors at the interfaces as the aim was to only wet the solid–solid interfaces. For each cell, the Nyquist plots (Fig. 2a) consisted of high frequency (5–0.1 MHz) and low frequency (kHz to Hz) semicircles that correspond to the bulk/grain-boundary and the Li|LLZO interface resistances, respectively. The resistance values were extracted from the x-axis intercepts of the plots. After introduction of the interlayers, we observed a reduction in the interface resistance by approximately a factor of 5, as the total resistance (RTotal) of the Li|LLZO|Li cells decreased from ∼6800 to ∼1350 Ω cm2 (Table 1). This suggests that the ion conducting IL gels fill up the micro voids at the solid–solid interfaces which were otherwise highly resistant to the ionic transport (Fig. S2, ESI†). Previous studies also showed the success of similar interlayer concepts in lowering the interface resistance and improve the Li+ ion transport.15,16,29
Li|LLZO|Li cell | Overpotential (V) | R Total at 25 °C (Ω cm2) | |||
---|---|---|---|---|---|
After 50 h at 0.1 mA cm−2 | After 200 h at 0.2 mA cm−2 | After 330 h | After 400 h at 0.4 mA cm−2 | ||
With interlayers | 0.12 | 0.21 | 0.31 at 0.3 mA cm−2 | 0.33 | 1350 |
Without interlayers | 0.67 | 1.30 | Short-circuited at 0.2 mA cm−2 | — | 6800 |
In the next step, the electrochemical stability of Li|LLZO interfaces with and without interlayers was tested by Li stripping/plating experiments. The comparison of the resulting voltage profiles for symmetrical Li|LLZO|Li cells is shown in Fig. 2b. The cells with interlayers displayed stable voltage profiles for more than 450 hours of continuous Li stripping/plating. The average overpotential values are listed in Table 1. Compared to Li|LLZO|Li with interlayers, a much higher overpotential was observed for cells without interlayers, as shown in the enlarged view in Fig. 2c. After 50 h of Li stripping/plating at 0.1 mA cm−2 the overpotential for the cells without interlayers was ∼0.67 V, while it was only ∼0.12 V with interlayers. Furthermore, the overpotential steadily increased with time for the cells without the interlayers. For example, at 0.2 mA cm−2 the overpotential increased from ∼1.2 V at 120 h to ∼1.60 V at 300 h, an indication of continuous deterioration of the interface during successive cycles. Such high overpotentials are likely to induce excessive stress on the Li|LLZO interface, thus initiating Li dendrites and accelerating their growth across the SE. This was evident from the sudden voltage drop at ∼330 h which suggests the onset of short-circuiting events, which will continue in the following cycles (Fig. 2d). This is consistent with the findings in other studies which reported adverse effects of an unstable Li|LLZO interface on the Li cyclability in the form of high overpotentials and Li dendrite formations.30,31 In contrast, the cells with interlayers show lower overpotentials, e.g., ∼0.31 V at 0.3 mA cm−2 after 330 h and ∼0.33 V after 400 h, and are able to withstand a current density as high as 0.4 mA cm−2. At all currents, the voltage profiles remained stable without noticeable deviations or irregularities, suggesting that the integrity of the interfaces was not affected. This may also suggest the formation of electronically insulating interphase layers that provide passivation and stop further decomposition of the electrolyte.32 Further, the lack of short-circuits indicates an effective suppression of the Li dendrite formation. Such improved interfacial properties of the system were possible by overcoming the physical and (electro-)chemical limitations imposed by the complex Li|LLZO interface. The IL based interlayers, firstly, filled up the micro voids at the solid–solid interface hence providing physical pathways for Li+ ion transport as illustrated in the schematic in Fig. S2.† Secondly, owing to the stability of Li with the PYR14TFSI ionic liquid,33 an electrochemically stable interphase layer was formed which facilitated the ionic transport. The chemical stability between LLZO and the IL gels was also investigated. The idea was to find out whether any chemical reactions occur between the chemical components of the interlayers and LLZO that could result in the formation of decomposition products. XRD analysis by comparing pristine LLZO and the one exposed to gels based on the IL, LiTFSI and organic solvents for one week showed similar diffraction patterns (Fig. S3, ESI†), indicating that at least no new crystalline phases were formed.
For full cell testing, sulfur–carbon composite cathodes were used. It is well known that S, lithium disulfide and other reaction intermediates have poor electronic conductivity.34 To improve this, S nanoparticles were impregnated into a nitrogen doped carbon matrix (from now on labelled ‘S/NC’); a detailed account on the synthesis of the cathode was given in our previous report.35 The carbon composite matrix with good electronic conductivity and a high surface area (700 m2 g−1) contained graphene and multiwall carbon nanotubes in an equal weight ratio. The morphology of the cathode material is illustrated in the SEM image shown in Fig. S4.† Further, thermo-gravimetric analysis (TGA) shows a S loading of ∼55 wt% in the S/NC cathode (Fig. S5, ESI†).
Fig. 3a shows the room temperature charge/discharge voltage profiles of a Li|LLZO|S/NC cell. The IL interlayers were introduced at both the anode and the cathode interfaces. The cells were cycled at a rate of 0.1C (1C = 1672 mA gsulfur−1) in a voltage window of 1.5–3 V. Two plateau voltage profiles (at ∼2.35 V, and ∼2.1 V) indicate the conversion based reduction reaction of S with Li, which is associated with the octahedral S8 ring opening and the formation of high- and low-order polysulfides (Li2Sx, x = 8–1), finally leading to lithium disulfide/lithium sulfide (Li2S2/Li2S). The discharge capacities for the 1st, 2nd, 50th, 100th cycles were ca. 1360, 1330, 712, and 570 mA h gsulfur−1, respectively. In the initial 30 cycles, noticeable capacity decay and a low coulombic efficiency (∼97%) were observed (Fig. 3b). In the following 70 cycles, the capacity stabilized and the coulombic efficiency increased to ∼99.8%. The initial lower coulombic efficiency may be attributed to the formation of the interphase layers on the Li anode and on the cathode side, resulting in an irreversible capacity loss.36 It is also evident in Fig. 3c that the voltage hysteresis increased from ∼300 mV to ∼500 mV in the initial 30 cycles, but stayed stable afterwards up to 100 cycles. We also tested the cell at 50 °C to check the impact of higher temperatures on the electrochemical performance. As shown in Fig. S6,† the capacity increased from ∼570 mA h gsulfur−1 (at 100th cycle, 25 °C) to ∼690 and ∼670 mA h gsulfur−1 for the next cycles (101th, and 102nd). The voltage hysteresis of the cell also reduced from ∼500 mV to ∼250 mV. This suggests that at higher temperatures better electrochemical performance may be achieved for the Li/S SSB, although in the current work our focus has been on room temperature performance.
Using LLZO ceramic as a separator is advantageous in physically blocking the polysulfide shuttle towards the Li anode. The higher order polysulfides (Li2Sx, x ≥ 4) created during the reduction reaction between the S based cathode and the IL based gel layers were confined to the S/NC|LLZO interface. As evident from the elemental maps in focused-ion beam (FIB) SEM images (Fig. 3d), LLZO SE is free from any S species. Nevertheless, the polysulfides, formed at the cathode interface may react with LLZO and create a more resistive interphase compared to the Li side. This prompted us to check the impedance contribution from each electrode separately. We compared the total impedances for Li|LLZO|S/NC full cells and Li|LLZO|Li symmetrical cells with interlayers after initial cycling as shown in Fig. S7.† A slightly higher impedance was observed for the Li|LLZO|S/NC cells which implies that the interphase layer formed at the S/NC|LLZO interface is more resistive than at the Li|LLZO interface.
In the next step, we analyzed the elemental composition and the chemical state (oxidation state) of the elements in the Li|LLZO and S/NC|LLZO interphase layers by XPS. For these measurements, we used the LLZO SE pellets after removal from the cycled electrochemical cells and subsequent washing with isopropanol. XPS spectra were recorded from both sides of the pellets, either that previously in contact with the Li anode or with the S/NC cathode side. Since XPS is a surface sensitive method, which probes only ∼3–5 nm of the outermost surface region, we acquired depth profiles of the elemental composition in the surface-near region by repeated XPS measurements after successive Ar+ ion sputter etching treatments. Overall, XPS spectra were recorded before sputtering and after sputtering times of 3, 7, and 20 minutes, resulting in total sputtering times of 0, 3, 10, and 30 min. Under the given conditions, the sputtering rate is expected to be ∼1 nm min−1. To get more information on the chemical state of the main elements in the interphase layer, the results of the detail spectra were analyzed by peak fits. Fig. 4 shows the spectra recorded on the Li|LLZO side, and the data of the S/NC|LLZO side are shown in Fig. S8.† The binding energies of the individual peaks which were used in the fit and the assigned species are also listed in Table 2.
Fig. 4 Detail XPS spectra of the solid interphase layer at the Li|LLZO interface. The spectra were recorded at the original surface and after 3, 10 and 30 minutes of Ar+ ion sputtering. |
Elements | Peak positions (eV) | Chemical species | |
---|---|---|---|
Lithium side | S/NC side | ||
C1s | 284.8 | C–C | C–C |
286.3 | C–O–C | C–O–C | |
288.4 | O–CO | O–CO | |
290.0 | Metal carbonates (e.g., Li2CO3) | Metal carbonates (e.g., Li2CO3) | |
293.0 | CF2/CF3 | CF2/CF3 | |
F1s | 685.0 | Metal fluorides (e.g., LiF) | Metal fluorides (e.g., LiF) |
688.7 | Organic fluorine (example –CF2/–CF3) | Organic fluorine (example –CF2/–CF3) | |
Li1s | 54.2 | Li in LLZO | Li in LLZO |
55.6 | Other Li species | Other Li species | |
N1s | 398.7 | Nitrides | — |
399.8 | TFSI− | TFSI− | |
402.4 | Pyr14+ | Pyr14+ | |
S2p | 168.8/170.0 | TFSI− | TFSI− |
167.0/168.2 | LixSOy | LixSOy | |
163.7/164.9 | Internal S of polysulfides | Internal S of polysulfides | |
161.9/163.1 | Disulfides (e.g., Li2S2)/terminal S of polysulfides | Disulfides (e.g., Li2S2)/terminal S of polysulfides | |
160.2/161.4 | Metal sulfides (e.g., Li2S) | Metal sulfides (e.g., Li2S) | |
O1s | 529.6 | LLZO/Li2O | LLZO/Li2O |
531.6 | CO | CO | |
Metal carbonates (e.g., Li2CO3) | Metal carbonates (e.g., Li2CO3) | ||
532.5 | C–O | C–O |
Fig. 5 displays the results of the elemental concentrations obtained for the two sides of LLZO electrolyte pellets as a function of the sputtering time. For the side in contact with the Li anode (Fig. 5b), the C content decreases sharply from 27 to 12 at% in the first sputtering step and is further reduced to 7 at% in the following steps. On the other hand, the concentrations of Li and O rise substantially upon sputtering and both arrive at close to ∼40 at% after the last sputtering step (removal of ∼30 nm material). At the same time, the F content increases initially (after 3 min of sputtering) but then decreases again, while the amount of N and S decreases slowly step-by-step upon sputtering. Finally, after the second sputtering step (removal of ∼10 nm material), small features of the metal ions in the Nb, Ba-doped LLZO (i.e., Zr, Nb, Ba and La denoted as ‘metals’ in Fig. 5) can be detected. Their intensity increases substantially in the spectra after the last sputtering step, leading to a total concentration of these four elements of ∼3 at%. These results indicate on the one hand that the interphase layer is at some points rather thin (∼10 nm). On the other hand, however, the persistence of clearly interphase-related elements such as F, together with the rather low concentration of the LLZO metals after the last sputtering step, shows that the interphase layer is of varying thickness. Further conclusions regarding the distribution of organic and inorganic species in the interphase layer will be drawn after the presentation and discussion of the XP spectra.
Fig. 5 Development of the elemental concentrations (at%) in the solid interphase layers at the (a) S/NC|LLZO and (b) Li|LLZO interfaces as a function of the total sputtering time. |
For the Li|LLZO side, the C1s detail spectrum of the original interphase surface, before sputter etching, was dominated by a peak at 284.8 eV, which is assigned to C–C/C–H groups. Further peaks at 286.4, 288.4, 290.0 and 293.0 eV are attributed to C–O, CO, (inorganic) carbonate (CO32−) and CFx species, respectively. In the spectrum recorded after the first sputtering step, the carbonate peak at 290.0 eV had by far the largest intensity, while that of all other peaks decreased substantially. Only the C–C/C–H peak at 284.8 eV still had a notable intensity. The spectra recorded after the subsequent sputtering steps showed essentially the same result. These finding are corroborated by the detail spectra of the O1s region. They show two peaks at 531.5 and 532.5 eV before sputtering, which are assigned to CO/metal carbonate and C–O species, respectively. After sputtering, the peak at 531.5 eV remained and dominated the O1s spectra. Additionally, another peak appeared in the spectra after sputtering at a lower binding energy (528.6 eV), which we relate to the presence of metal oxide species (Li2O and LLZO). Going to the F1s region, the spectra recorded before sputtering showed two main peaks at 685.0 and 688.6 eV, which we attribute to metal fluorides (i.e., LiF) and CFx groups (in TFSI− or its decomposition products), respectively. After sputtering the peak assigned to CFx functional groups disappeared completely, while the metal fluoride peak at a lower binding energy persisted. The S2p spectrum recorded before sputtering is dominated by two peak doublets, which we assign to TFSI− (169.3/170.5 eV) and to oxidized decomposition products of it (e.g., LixSOy; 167.1/168.3 eV). After sputtering, new features appeared at a lower binding energy, which we relate to (metal) sulfide species, and which grew in intensity with increasing sputtering time. Next, for the measurements in the N1s region, before sputtering a dominant peak at 399.8 eV, which is related to the N atom of the TFSI− anion, and a much smaller feature at 402.4 eV due to the N atom in the Pyr14+ cation of the ionic liquid were shown. After sputtering, a new peak was detected at 398.7 eV, which we assign to nitride species. Finally, it may be noted that the detail spectra of the LLZO constituents (Zr3d, La3d, etc.) appearing after the last sputtering step (Fig. S9, ESI†) were in all cases dominated by features of fully oxidized species.37 In the case of Zr and Nb, there were in addition small contributions from reduced species, which are most probably a result of sputter-induced reduction processes.
The spectra and the elemental concentrations obtained for the side of the electrode, which was in contact with the S/NC cathode, showed in general similar features and trends as were observed on the Li side (Fig. 5a and S8, ESI†). For example, the spectra in the O1s and F1s region showed before sputtering a very similar result with features related to C–O and CO/metal carbonate (O1s region) species or metal fluoride and CFx species, respectively, while after sputtering the CO/carbonate (O1s region) and metal fluoride (F1s region) peaks were observed again (together with a small additional feature due to metal oxide species in the O1s region). Still, slight differences could also be noted. For example, the spectra in the C1s region showed also an increase in the peak related to metal carbonates (at 290.0 eV) upon sputtering, though the C–C/C–H peak remained a major contribution in the spectra of this side. The peak positions and associated chemical species at the Li|LLZO side and S|LLZO side are summarized in Table 2.
To provide further evidence of the presence of interphase layers, we carried out STEM analysis of the cycled LLZO particles that was interfaced to the Li anode. The STEM-HAADF image reveals the presence of interphase layers around the LLZO particles (Fig. 6a). Elemental mapping obtained through EDS spectrum imaging shows that the interphase is rich in S, F, C and O species while LLZO has a dominant concentration of La and Zr. S and F follow the same spatial distribution inside the interphase with minor differences registered in the distribution of O (Fig. S10, ESI†). The cumulated EDS spectra presented in Fig. 6b show that the interphase region has a chemical composition dominated by C, S, F and O that is substantially different from that of the LLZO particles. Note that Li is undetectable using conventional EDS detectors. The HRTEM image in Fig. 6c also reveals an amorphous interphase layer on the LLZO particles with variable thickness (range ca. 5–15 nm) as highlighted by the dashed lines. This agrees well with the XPS analysis where the metal ions (i.e., Zr, Nb, Ba and La) in LLZO were detected only after the second sputtering step (removal of ∼10 nm material). Inspection of the selected region of LLZO reveals the lattice fringes with a d spacing of 3.45 Å (inset Fig. 6c), which corresponds to the (321) hkl plane of the cubic LLZO phase. Moreover, the fast Fourier transform (FFT) images in Fig. 6c and the selected are electron diffraction patterns shown in Fig. S11† prove the cubic structure of the LLZO phase and its high crystallinity. In a higher magnification HRTEM image (Fig. S12, ESI†), another lattice spacing d = 2.09 Å of the (611) plane of cubic LLZO appears.
Overall, XPS and STEM analyses confirm the presence of amorphous solid interphase layers on the LLZO surface which are rich in S, F, C and O species and in the nano-meter range. The interphase layers are of varying thickness. A significant decrease in the carbon content upon sputtering, together with the clear increase of the content of inorganic (metal) carbonate and fluoride species is observed. The development of the elemental concentrations demonstrates that the outermost part of the interphase layer contains mostly organic species, while the inner part (close to the LLZO surface) mainly consists of inorganic species (cf. schematic depiction Fig. 7).
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0ta05014h |
‡ These authors contributed equally. |
This journal is © The Royal Society of Chemistry 2020 |