Journal of Electroceramics https://doi.org/10.1007/s10832-018-0129-y

Importance of mixing protocol for enhanced performance of composite cathodes in all-solid-state batteries using sulfide solid electrolyte Sungwoo Noh 1 & William T. Nichols 1 & Moonju Cho 1 & Dongwook Shin 1 Received: 10 January 2018 / Accepted: 28 February 2018 # Springer Science+Business Media, LLC, part of Springer Nature 2018

Abstract All-solid-state battery performance is strongly dependent on effective charge transfer at both 1) the interface of the active particles and 2) through the interstitial regions of composite cathode. Design of the composite cathode is further complicated by the necessity to limit the amount of conductor additives in order to attain high energy density. These requirements present a difficult design challenge for the composite cathode. Here we investigate the extent to which the mixing order of the three components in the composite cathode impacts the charge transfer and cell performance. We test a total of 5 mixing protocols and find that the initial discharge capacity and the rate capability varies significantly with mixing order. It is shown that the location of the electron conductive carbon is particularly critical for cell performance due to its limited quantity in the composite cathode. Mixing protocols that concentrate the carbon at the active particle interface lowers the interfacial resistance leading to higher discharge capacity. Mixing protocols that place more carbon in the interstitial regions improves the electron path conductivity and is found to correlate with higher rate capability. Based on these results we demonstrate a mixing protocol that achieves both higher discharge capacity and better rate performance for all-solid-state batteries. Keywords All-solid-state battery . Solid electrolyte . Conductive carbon . Composite cathode

1 Introduction Lithium ion batteries are common in portable electronic devices because of their high energy density and long cycle life. Recently, large-scale lithium-ion batteries for electric vehicles have attracted much attention, but commercialization has been hampered due to concerns about the safety of batteries employing conventional organic liquid electrolytes [1]. The safety issues of the liquid electrolyte are primarily caused by chemical reactions with the active cathode materials at elevated temperature, electrolyte leakage and a narrow electrochemical window [2, 3]. As a result, considerable efforts have been focused on developing all-solid-state lithium ion batteries [4–6]. For all-solid-state lithium ion batteries, solid electrolytes with a high ionic conductivity are required to achieve good electrochemical performances. Among solid electrolytes, sulfide-based solid electrolytes have shown significant * Dongwook Shin [email protected] 1

Division of Materials Science & Engineering, Hanyang University, 222 Wangsimni-ro, Seongdong-gu, Seoul 133-791, South Korea

promise because of good lithium ion conductivity as well as a wide electrochemical window [7–11]. In all-solid-state batteries, the cathode composite consists of three component powders: active cathode, solid electrolyte and electron conductive carbon. The composite cathode must simultaneously provide 1) interfacial transfer of both electrons and lithium ions across the surface of the active particles and 2) conductive pathways for both electrons and lithium ions through the composite cathode as suggested in Fig. 1. Furthermore, in order to attain high energy density, it is necessary to limit the amount of conductor additives in the cathode, which puts tight constraints on the composite cathode design. An effective way to address this challenge is through careful dispersion of the conductive powders within the composite cathode through the mixing process. In our previous report, it was noted that the microstructure and morphological properties of the resulting composite depended on the conditions of wet mixing in a ball mill [12]. This led us to expect that the composite cathode performance could be improved through careful attention to the mixing order of the component powders. In this paper, we examine if the mixing order of the three components in the composite cathode effects the cell performance. We test a total of 5 methods and find the initial

J Electroceram

Fig. 1 Schematic showing the internal structure of the composite cathode within an all-solid-state battery cell. The key design requirements are illustrated: (a) Path resistances through the composite cathode must be small for both electrons and lithium ions to enable sufficient transport to their respective current collectors. (b) Interfacial resistances for both

electrons and lithium ions must also be small to enable effective lithiation and delithiation of the active particles. To satisfy these requirements carbon and solid electrolyte particles must be in intimate contact with the active particles and also form continuous conducting pathways through the composite cathode

discharge capacity and the rate capability varies significantly with mixing order. We find that the location of the conductive carbon is particularly critical for good cell performance due to its limited quantity in the composite cathode. Mixing protocols that concentrate the carbon at the active particle interface lowers the interfacial resistance leading to higher discharge capacity. Mixing protocols that place more carbon in the interstitial regions improves the electron path conductivity and is found to correlate with higher rate capability. Based on these results, we demonstrate that mixing half of the electron conductor with the solid electrolyte and reserving the other half for the final mixing produces both a high discharge capacity and the best rate performance.

of 16 mm diameter. Both faces of the pellets were then attached to stainless steel disks as current collectors. To fabricate all-solid-state cells, the 96(78Li2S·22P2S5)· 4Li 2 SO 4 glass-ceramics powders and indium foil (99.9975%, ALFA AESAR) were used as solid electrolytes and the anode layers, respectively. Prior to preparing the cells, the surface of the LiCoO2 active material was coated with Li 2 CO 3 using lithium hydroxide solution via lowtemperature heat treatment [17]. The composite electrodes were prepared by mixing the surface modified LiCoO 2 , 96(78Li 2 S·22P 2 S 5 )·4Li 2 SO 4 glass-ceramic and Super P® carbon (Timcal) at a weight ratio of 70:28:2 using a mortar and pestle with 5 different orders of mixing as summarized in Fig. 2. The weight ratio of the composite cathode was selected based on our previous study that showed it provided the highest active material percentage while still maintaining the capacity [18]. All other mixing parameters such as the mortar used, and mixing time were the same for all methods. The microstructures of the composite electrodes prepared by the five mixing methods were examined to identify the change in morphologies and how the constituent particles were distributed using a field emission scanning electron microscope (FE-SEM, S-4800, Hitachi, Japan) equipped with energy-dispersive X-ray spectroscopy (EDX) for determining the elemental distribution. DC polarization of the composite electrodes with different mixing order was measured to determine a lithium ion resistance and electron resistance of the samples. To make the measurements, 400 mg of the composite cathode powder mixed by each of the five methods were cold-pressed under 4 metric tons in a 16 mm diameter mold to prepare a 1 mm composite pellet. Lithium plates and stainless steel (SUS)

2 Experimental procedure The solid electrolyte used in this work is the 96(78Li2S· 22P2S5)·4Li2SO4 glass-ceramic prepared using a previously reported mechanical milling process with subsequent heat treatment [13–16]. Reagent grade Li2S (99.9% purity, Alfa Aesar), P 2S 5 (99% purity, Sigma-Aldrich) and Li 2 SO 4 (99.99%, Sigma-Aldrich) were mixed thoroughly in the appropriate molar ratios. Then mechanical milling was performed at 520 rpm for 20 h using a high-energy planetary ball mill (Pulverisette 7, Fritsch) with milling cycles of 40 min followed by resting for 20 min. Lithium ion conductivity measurements of the solid electrolyte were carried out with an electrochemical impedance analyzer (Wonatech SP5) operated in the frequency range of 0.1 Hz ~1 MHz. For these measurements, the powders were pelletized under 4 metric tons pressure with an uniaxial press

J Electroceram

Fig. 2 Protocols for the five different mixing methods along with the anticipated microstructures of the resulting composite cathodes

plates were attached onto each face of the pelletized sample as non-blocking electrodes and blocking electrodes, respectively. Time dependence of the electrical current was measured under constant voltage 1 V for 3600 s. To make the complete cells, 20 mg of the composite cathode mixture, 200 mg of the 96(78Li2S·22P2S5)·4Li2SO4 glass-ceramics powder and an indium foil (0.1 mm) were cold-pressed together under 4 metric tons in a 16 mm diameter mold. The obtained three-layer pellets were packaged in a 2032-type coin cell for electrochemical measurements. All processes were performed in a dry, Ar-filled glove box ([H2O] <1 ppm) [19]. All cells were galvanostatically charged and discharged using a charge–discharge measurement device (TOSCAT3100, Toyo System) at room temperature. The charge–discharge performance was evaluated with cut off voltages of 1.88–3.68 V (vs. Li–In) with 0.05 C to 1.0 C current densities. Electrochemical impedance spectroscopy measurements of the obtained cells were performed using an impedance analyzer (Wonatech mp5) in the frequency range from 0.1 Hz to 1 MHz after charging to 3.68 V.

3 Result and discussion The objective of this work is to understand how the mixing order effects the discharge capacity and rate capabilities of composite cathodes. It is expected that first mixing two powders together will bring them into intimate contact promoting interfacial conductivity. In contrast, subsequently mixing in a third powder will have less interfacial contact, but a larger presence in the interstitial regions promoting higher path conductivity. Based on these considerations we have chosen five possible orders to mix the three powder components as given

in Fig. 2 along with the expected microstructure of the composite cathodes. In method 1 (M1), mixing the active particles first with the conductive carbon would be expected to enhance the electron transfer across the cathode interface as suggested in Fig. 2(a). In method 2 (M2), the active particles are first mixed with the solid electrolyte to promote the lithium ion transfer across the cathode interface. The electron conductive carbon is subsequently mixed in where it is expected to provide higher interstitial conductivity within the composite cathode as suggested in Fig. 2(b). Method 3 (M3) Fig. 2(c), mixes of all three powders simultaneously. This is the standard protocol for making composite cathodes and aims at a compromise between lithium ion and electron transfer across the cathode particle interface. A drawback of M3, however, is that the solid electrolyte and conductive carbon powders might not coat the cathode particles in optimum quantities. To address this limitation, method 4 (M4) first mixes the solid electrolyte and conductive carbon together creating a pre-mixed conductor powder, then this powder is subsequently mixed with the active particles as shown in Fig. 2(d). From our analysis of the composite cathodes from M1 to M4, an improved method was suggested as will be discussed below. In method 5 (M5), the preparation is similar to M4, except only half of the electron conductive carbon is mixed initially, then the second half is mixed in at the end, as shown in Fig. 2(e). This was found to be advantageous to promote both electron transfer across the cathode interface and electron transport through the active composite as will be discussed below. To test our model, scanning electron microscopy with EDX point analysis are taken from selected areas within the crosssection of the composite cathodes as shown in Fig. 3. In each sample, the initial analysis region is at the LiCoO2 active particle surface and the following two analysis region are measured at increasing distance into the interstitial region

J Electroceram Fig. 3 Cross-section SEM images and EDX point analysis of 9 selected areas with composite cathodes prepared by (a) Method 1, (b) Method 2 and (c) Method 4

between active particles. The calculated elemental composition results are shown in the adjoining table in Fig. 3. In sample M1 the percentage of carbon decreases as we move away from the surface and the percentage of the solid electrolyte increases (points ①, ② and ③ in Fig. 3(a)). In contrast, M2 shows that the percentage of carbon increases with increasing distance from the active particle surface. These results confirm our expectations in Fig. 2 that directly mixing with the active material concentrates carbon at the surface (method 1) whereas subsequently mixing at the final step concentrates the carbon in the interstitial regions between particles (method 2). In M4 the carbon and solid electrolyte are first mixed in order to make a uniform dispersion of the two conductors prior to mixing with the active particles. We find that the carbon mass percentage is relatively uniform from the surface out into the interstitial region between active particle. These results again provide support for our simple model of the expected microstructure for different mixing orders suggested in Fig. 2. With confidence in our microstructure model we can evaluate the performance of the composite electrodes. Figure 4 shows the first cycle discharge capacities. We find an increase

from 78.4 mAh/g for sample M1 to 150.5 mAh/g for sample M4 as summarized in Table 1. The poor capacity of sample

Fig. 4 Charge-discharge curves of all-solid-state cells made from mixing methods M1-M4. Measurements were made at room temperature at 0.05C charge and discharge rate with cut off voltages of 1.9–3.68 V (vs. Li–In). (A: cathode material, S: solid electrolyte, C: conductive carbon)

J Electroceram Table 1 Summary of the 0.05C discharge capacity, 1C discharge capacity, charge-transfer resistance, lithium-ion path resistance and electron path resistance for samples M1-M5 Discharge Conditions Capacity

Method 1 Method 2 Method 3 Method 4 Method 5

0.05C (mAh/g)

1C (mAh/g)

78.4 109.6 137.9 150.5 144.0

13.1 38.6 28.0 32.4 54.2

ChargeLithium ion Electron Transfer R path R path (Ω) (Ω/100 μm) R (kΩ/ 100 μm)

54.9 47.4 41.1 31.2 37.3

9.2 11.8 13.8 11.8 14.2

5.2 × 102 1.6 × 102 5.4 × 102 5.8 × 102 11

M1, with the carbon mixed directly with the acitive particles, suggests that carbon particles may be preventing interfacial contact of the solid electrolyte to the active particles limiting the lithium ion transfer. With the solid electrolyte at higher concentrations at the surface (M2) there is a definite improvement to 109.6 mAh/g. However, even better results were obtained when a combination of both the solid electrolyte and electron conducting carbon at the surface; 137.9 mAh/g for M3 and 150.5 mAh/g for M4. Clearly, this points to the need to have both carbon and solid electrolyte at the active particle surface for effective interfacial transport of electrons and lithium ions. The difference between M3 and M4 is also instructive. In sample M4, the solid electrolyte and carbon are first pre-mixed together creating an intimate dispersion of both electron and ion conductors that can uniformly coat the cathode particles as suggested in Fig. 2(d) and confirmed in Fig. 3(c). On the other hand, mixing all three components at once

(M3) appears to lead to an uneven dispersion of each conductor at the surface resulting in about 10% lower capacity. The composite cathode microstructure must also provide a high rate capacity and have robust cyclical performance for applications. Figure 5 shows the rate capabilities of the samples and the values at a rate of 1C (cycle N = 21) are summarized in Table 1. While the samples M3 and M4 have a high initial capacitance at low rates (0.05 C), their capacities fall off more dramatically to 28.0–32.4 mAh/g for a high 1 C discharge rate. On the other hand, M2 maintains its capacity much more effectively with a value of 38.6 mAh/g at the 1 C rate. Taking into consideration both the results of the initial discharge capacities (Fig. 4) and the rate performance (Fig. 5) clearly suggests that there are two distinct design issues. First, for high discharge capacity, it is necessary for both the solid electrolyte and electron conductive carbon to have intimate contact with the active particle surface as discussed previously. Second, for improved cyclical and high rate performance, it is better to mix the conductive carbon later in order to incorporate sufficient quantities into the interstitial regions as suggested in Figs. 2(b) and 5. Thus, we would predict that mixing the conductive carbon both initially with the active particles to provide interfacial contact and subsequently to provide carbon in the interstitial regions for higher electron pathway conductivity. To accomplish this, sample M5 was prepared using a combination of the M2 and M4 mixing protocols. In the first mixing, the solid electrolyte and only half of the conductive carbon were mixed producing a pre-mixture of conductors that are subsequently mixed with the active particles. Then as the final step, the other half of the conductive carbon was mixed in. In this way the carbon is expected to be present at both the active particle surface to provide a higher interfacial conductivity and in the interstitial regions to provide higher pathway conductivity. To investigate the effectiveness of this

Li / Sample / Li

Current (A)

0.01 1E-3 1E-4

SUS / Sample / SUS

1E-5 1E-6 1E-7 0.0

Fig. 5 Rate capability of all-solid-state cells made from mixing methods M1-M4. Measurements were made at room temperature at 0.05C to 1C discharge rate cut off voltages of 1.9–3.68 V (vs. Li–In). (A: cathode material, S: solid electrolyte, C: conductive carbon)

M1 M2 M3 M4 M5

0.2

0.4

0.6

0.8

1.0

Time (h) Fig. 6 DC polarization curves for the composite cathode samples M1-M5 measuring ion conduction (Li / sample / Li) and electron conduction (stainless steel / sample / stainless steel)

J Electroceram

approach, DC polarization and AC impedance measurements can provide information on the pathway resistance and interfacial resistances respectively. DC polarization measurements allow the total current flow of electrons and lithium ions through the composite cathode to be determined. In order to differentiate between the ionic and the electronic current, the DC polarization curves of each composite cathode sample are measured using lithium ion nonblocking electrodes (Lithium metal) for ionic current and electrical current at the same time and electron non-blocking electrodes (stainless steel) for electrical current as shown in Fig. 6. Within the first few minutes the current of all samples reaches a steady state. The DC polarization curves of non-blocking electrodes can be considered as ionic currents because of DC polarization curves of blocking electrodes are too low compared to non-blocking electrodes. All five composite cathode samples with different mixing order have similar ionic current. These results are expected because of the relatively higher solid electrolyte content (28% wt) compared to the conductive carbon (2% wt). This emphasizes that the location of the carbon conductor particles is critical in the performance of the overall composite cathode. The electron current is particularly low for samples M1, M3 and M4. In each case carbon was mixed first, resulting in relatively little carbon within the interstitial regions of the composite matrix. Jamnik et al. showed that electron transport typically occurs via particle-particle contact (carbon/ carbon, carbon/cathode, and cathode/cathode) [20]. Since the amount of conductive carbon is small, this leads to tortuous electron pathways through the composite cathode, increasing the electron path resistance. In sample M2 on the other hand, the conductive carbon was mixed at the end, resulting in a more connected network of carbon particles throughout the composite cathode resulting in a lower electron resistance. Figure 6 shows that this results in an order of magnitude increase in the electron transport. As was seen in Fig. 5, sample M2 also displayed the best rate performance suggesting that electron transport is the limiting factor in the composite cathode at high discharge rates. Despite the good rate performance, however, sample M2 did have a lower initial discharge capacity at lower rates. This is attributed to a reduced amount of conductive carbon at the interface of the cathode particles limiting the amount of charge transfer across the interface. For sample M5 the electron path current increases by more than an two orders of magnitude compared to the other samples. The electron conductivity is also higher than sample M2 where a majority of the conductive carbon was in the interstitial region between the cathode particles. This can be understood because in sample M5 the conductive carbon particles are also present at higher concentration at the interface with the active particles in addition to the interstitial region. Thus, the mixing procedure for sample M5 ensures effective charge transfer at the interface and continuous electron conduction pathways through the interstitial regions providing faster charging/discharging.

To understand the importance of interfacial resistance the AC electrochemical impedance of the all-solid-state cells were measured as shown in Fig. 7. In general, the impedance components overlap in the spectra making it difficult to give precise values of the different impedance components. Nevertheless, each of the mixing order samples displays notable differences in the sizes of the semicircles in the Nyquist plot, therefore, fitting can give a reasonable estimate of the interfacial impedance. Sakuda et al. analyzed impedance plots for In/SE/LiCoO2 cells and identified 3 resistance components. The high-frequency region (>100 kHz) was attributed to the solid electrolyte resistance (RSE), the intermediatefrequency region (~ 500 Hz) to the cathode interface resistance (RPE) and the low-frequency region (~ 1 Hz) to the LiIn electrode resistance (RNE) [21]. Figure 7 shows the impedance plots for In/SE/LiCoO2 cells after charging to 3.68 V vs Li-In during the first cycle. The figure clearly shows that the mixing order strongly affects impedance components. Fitting the impedance plots using Sakuda et al.’s assignment, the cathode interface resistance RPE of sample M1 to M5 follow the same trend as the discharge capacity as shown in Table 1.

Fig. 7 AC impedance spectra of all-solid-state cells made with mixing methods M1-M5. Measurements were made at room temperature after full charging to 3.68 V

J Electroceram

the carbon at the active particle interface, lowers the interfacial resistance leading to higher discharge capacity. Further, mixing protocols that place more carbon in the interstitial regions between active particles improves the electron path conductivity and is found to correlate with higher rate capability. Using these results we demonstrtated that mixing half carbon electron conductor with the solid electrolyte with the other half reserved for the final mixing stimultaneously achieved both initial capacity and rate performance. These results suggest the need for careful design of the sysnthesis protocol for composite cathodes in all-solid-state batteries.

Fig. 8 Rate capability of all-solid-state cells made with methods M2, M4 and M5. Measurements were made at room temperature at 0.05C to 1C discharge rate with cut off voltages of 1.9–3.68 V (vs. Li–In). (A: cathode material, S: solid electrolyte, C: conductive carbon)

Thus, we see that for sample M3, M4 and M5 mixing protocols, both carbon and solid electrolyte are present at the active particle’s surface leading to a lower interfacial resistance and as a result higher initial capacity. From the results of the path conductivity (Fig. 6) and interfacial conductivity (Fig. 7) we see that there is a trade-off between mixing the limited amount of carbon particles into the interfacial regions (M2) or mix more carbon to the surface of the active particles (M4). Sample M5 was designed to provide a compromise that aims to achieve both higher pathway and interfacial resistance. Figure 8 compares the rate performance of the samples M2, M4, and M5. Due to the slightly higher interfacial resistance (Table 1, Fig. 7), the initial discharge capacity of sample M5 is about 5% lower than M4. On the other hand, by virtue of the much higher electron conductivity through the interstitial regions, the capacity at the high 0.5C and 1 C rates are dramatically better for M5 at 88.6 and 54.2 mAh/g respectively. From these results we see that the synthesis protocol of composite cathodes should assure that the solid electrolyte and carbon particles are both present at interface with the active particles for high transfer rates and that carbon particles are well dispersed within the interstitial regions to allow the carriers to move throughout the cathode.

Acknowledgements This work was supported by the Dual Use Technology Program of the Institute of Civil Military Technology Cooperation granted financial resources from the Ministry of Trade, Industry & Energy and Defense Acquisition Program Administration (17-CM-EN-11).

References 1. 2. 3. 4. 5.

6. 7. 8. 9. 10.

11. 12. 13. 14. 15.

4 Conclusions In this report we have shown that the initial discharge capacity and the rate capability varies significantly with mixing order of the components of the composite cathode. Due to the very small amount of conductive carbon (2% wt), its location within the composite cathode was found to be particularly critical for good cell performance. Among the 5 mixing protocols tested, we observed that mixing protocols that concentrate

16. 17. 18. 19. 20. 21.

M. Armand, J.M. Tarascon, Nature 451, 652–657 (2008) J. Kalhoff, G.G. Esthete, D. Bresser, S. Passerine, ChemSusChem 8, 2154–2175 (2015) Y. Wang, W.H. Zhong, ChemElectroChem 2, 22–36 (2015) J.B. Bates, N.J. Dudney, G.R. Gruzalski, R.A. Zuhr, A. Choudhury, C.F. Luck, J.D. Robertson, Solid State Ionics 53, 647–654 (1992) J.G. Kim, B. Son, S. Mukherjee, N. Schubert, A. Bates, O. Kwon, M.J. Choi, H.Y. Chung, S. Park, J. Power Sources 282, 299–322 (2015) E. Quartarone, P. Mustarelli, Chem. Soc. Rev. 40, 2525–2540 (2011) J.L. Souquet, E. Robinel, B. Barrau, M. Ribes, Solid State Ion. 3-4, 317–321 (1981) J.P. Malugani, G. Robert, Solid State Ionics 1, 519–523 (1980) A. Hayashi, S. Hama, F. Mizuno, K. Tadanaga, T. Minami, M. Tatsumisago, Solid State Ionics 175, 683–686 (2004) N. Kamaya, K. Homma, Y. Yamakawa, M. Hirayama, R. Kanno, M. Yonemura, T. Kamiyama, Y. Kato, S. Hama, K. Kawamoto, A. Mitsui, Nat. Mater. 10, 682–686 (2011) Y. Seino, T. Ota, K. Takada, A. Hayashi, M. Tatsumisago, Energy Environ. Sci. 7, 627–631 (2014) J. Kim, M. Eom, S. Noh, D. Shin, J. Power Sources 244, 476–481 (2013) S. Choi, M. Eom, C. Park, S. Son, G. Lee, D. Shin, Ceram. Int. 42, 6738–6742 (2016) J. Kim, M. Eom, S. Noh, D. Shin, Electron. Mater. Lett. 8, 209–213 (2012) M. Tatsumisago, S. Hama, A. Hayashi, H. Morimoto, T. Minami, Solid State Ionics 154, 635–640 (2002) M. Eom, J. Kim, S. Noh, D. Shin, J. Power Sources 284, 44–48 (2015) J. Kim, M. Kim, S. Noh, G. Lee, D. Shin, Ceram. Int. 42, 2140– 2146 (2016) S. Noh, W.T. Nichols, C. Park, D. Shin, Ceram. Int. 43, 15952– 15958 (2017) S. Noh, J. Kim, M. Eom, D. Shin, Ceram. Int. 39, 8453–8458 (2013) J. Jamnik, G. Miran, MRS Bull. 34, 942–948 (2009) A. Sakuda, H. Kitaura, A. Hayashi, K. Tadanaga, M. Tatsumisago, J. Electrochem. Soc. 156, A27–A32 (2009)