Evolution of Solid Electrolyte Interphase during Cycling and Its Effect on Electrochemical Properties of LiMn 2 O 4

Thickness variation of the solid electrolyte interphase (SEI) produced during charge-discharge cycling is investigated to analyze the effect of SEI on the electrochemical properties of LiMn 2 O 4 . Atomic force microscopy (AFM) is used to measure the SEI thickness and elastic modulus on the LiMn 2 O 4 surface. The SEI shows a broad thickness distribution due to the random nature of the LiMn 2 O 4 electrode surfaces, while the average SEI thickness increases with cycling and stabilizes after the 20 th cycle. Formation of a relatively thin SEI on the LiMn 2 O 4 surface accompanies low Coulombic efﬁciency at early cycling stages. The SEI produced in the early stages of cycling is vulnerable to capacity fading due to inefﬁcient surface protection against possible side reactions. A fully-grown stable SEI after 20 cycles shields the cathode surface from the electrolyte, minimizing capacity fading.

Solid electrolyte interphase (SEI) is the layer produced on the surface of active materials during charge-discharge cycling of a battery. SEI affects electrochemical properties of active materials, such as capacity retention and rate performance, and a stable SEI is known to protect the active materials from electrolyte side reactions. A wellknown example is the excellent cycling performance of graphite due to the thin SEI produced on the surface. 1,2 SEI control for cathode materials is, therefore, important to improve the electrochemical performance of Li-secondary batteries. In particular, much interest has been given to the study of SEI formation on LiMn 2 O 4 particles, as its SEI is relatively unstable and ineffective in protecting the electrode surface from side reactions with a protic electrolyte, causing capacity fading due to Mn 2+ ion dissolution. [3][4][5] Although LiMn 2 O 4 has been considered as a promising cathode material for lithium ion secondary batteries due to its stability and high discharge voltage, as well as its non-toxicity and low cost, its capacity fading is known to be a part of the reason for its limited commercial application.
Many researchers have studied the effect of SEI on electrochemical properties; however, the transient nature of SEI formed during cycling, and their nanoscale dimension made detailed analyses difficult. 6,7 Transmission electron microscopy (TEM) has been used for the observation of SEI, 8,9 however, this has inherent shortcomings since SEI, which is composed of organic and inorganic materials (LiF, Li 2 CO 3 , R-CO 3 Li), can be decomposed at ultra-high vacuum conditions and by the high energy focused electron beam. 10 Other techniques, such as spectroscopic ellipsometry 11 and X-ray reflectivity analysis, 12 were also used to analyze SEI. They found that the thickness of the SEI layer was less than 5 nm in the early stages of cycling. [10][11][12] Atomic force microscopy (AFM) has been used to examine electrode surfaces, while most AFM research has been focused on the change in topography, particle morphology, and electrical conductivity of the electrode surface. [13][14][15] Recently, Zhang et al. found that the SEI on a MnO anode surface is inhomogeneous with broad distributions of both the thickness and mechanical properties. 16 In this study, the SEI produced on the LiMn 2 O 4 particles in a cathode during cycling is studied by using AFM. Particular attention was given to SEI evolution during cycling concerning the thickness and elastic modulus and the effects of SEI on the electrochemical properties of LiMn 2 O 4 are discussed.

Experimental
A LiMn 2 O 4 thin-film was prepared by a spin coating method to produce a binder-free electrode with a smooth surface for AFM analysis. Lithium acetate dihydrate (0.012 mol) and manganese acetate z E-mail: hojang@korea.ac.kr tetrahydrate (0.024 mol) with a stoichiometric composition were dissolved in 2-mothoxyethanol (47 ml) and of ethanolamine (3 ml). The solution was coated on a stainless steel plate (1.86 cm 2 ) by spin coating (3000 rpm for 10 s), and the precursor coated substrate was heated at 350 • C for 10 min. The spin coating, followed by heating, was repeated 20 times. The final LiMn 2 O 4 thin-film electrode was obtained by annealing the precursor-coated substrate at 600 • C for 2 h in air.
For the battery cycling test, 2032 coin type cells were assembled using the LiMn 2 O 4 thin-film as a working electrode. Li metal, Celgard2500, and 1 M LiPF 6 dissolved solution of ethylene carbonate (EC), dimethyl carbonate (DMC), and ethyl methyl carbonate (EMC) (EC/DMC/EMC = 1:1:1 volume ratio) were used as a counter electrode, separator, and electrolyte, respectively. The separator was soaked in the electrolyte for 24 h before cell assembly. Cycling tests were performed with a VMP3 (Bio-Logic) with a current density of 50 μA/cm 2 over a voltage range of 3.5-4.5 V. After the cycling tests, the electrodes were rinsed using DMC and dried at room temperature in an argon-filled glove box to remove residual LiPF 6 on the electrode surface before subsequent AFM analysis. 3 An AFM (XE-100, Park Systems) was used for SEI analysis before and after the cycling test. AFM measurements were carried out with a cantilever with a stiffness of 42 nN/nm. Topography was obtained in a non-contact mode to avoid SEI destruction by the AFM tip (tip radius ∼ 10 nm). Nano-indentation tests (indenting speed ∼ 50 nm/s) were performed with the same cantilever to examine the surface layer thickness in randomly selected locations on the electrode surface. Indentation experiments using the AFM were carried out in air within 20 min to minimize possible contamination or degradation.
The crystal structure and morphology of the LiMn 2 O 4 particles were examined by X-ray diffraction (XRD, D/MAX-II A, Rigaku, using Cu K α radiation) and scanning electron microscopy (SEM, FE-SEM, Hitachi, S-4300).

Results and Discussion
The surface morphology and composition of the thin-film electrodes were analyzed before cycling tests and subsequent nanoindentation tests. Fig. 1 shows diffraction peaks obtained from the thin-film electrode produced on a steel substrate, indicating that the thin-film is LiMn 2 O 4 with a spinel structure. Some peaks corresponding to the Mn 2 O 3 impurity phase were also observed, which is known to be found, due to volatile lithium, in the case of LiMn 2 O 4 thin-film electrodes prepared by spin coating 17 or pulsed laser deposition. 18 However, the voltage plateau of Mn 2 O 3 (lower than 3 V vs. Li + /Li) is quite different with that of LiMn 2 O 4 , 19 thus the effect of Mn 2 O 3 on electrochemical properties of thin film electrode is not significant. AFM examination indicated that the thin-film was composed of nanosize particles less than ∼100 nm in diameter, and that the average surface roughness (R a ) was approximately 4.5 nm, which is similar to the LiMn 2 O 4 thin-film electrodes reported by other researchers (Fig. 2). 20,21 Particles were not facetted, but rounded, suggesting that LiMn 2 O 4 nanoparticles grew without maintaining the preferred lowenergy crystal planes. Figure 3 shows the electrochemical performance of the LiMn 2 O 4 thin-film electrode up to 40 cycles at a current density of 50 μAh/cm 2 . A voltage plateau was found at approximately 4 V (an inset of Fig. 3), which is consistent to the voltage plateaus reported by other researchers. 3,6,7 Coulombic efficiency at the initial cycle was relatively low (72%) and it rapidly increased at the 2 nd cycle, and stabilized to 98-99% after the 20 th cycle. On the other hand, the capacity retention decreased drastically to 88% during the early 10 cycles and it was maintained at approximately 81∼84% after 20 th cycle. These results suggest that the electrochemical property of the thin film electrode was stabilized after 20 th cycle, while the Coulombic efficiency and capacity fading were significantly changed during early cycles.
In order to investigate the effect of the SEI on the electrochemical performance, the electrodes, after different cycle tests (1, 5, 10, 20, and 40 charge-discharge cycles), were retrieved from the cell for Before the contact with electrolyte, soft SEI was not found on the surface of the LiMn 2 O 4 thin film: only a hard substrate was detected (Fig. 4b). On the other hand, a soft SEI was detected on the cycled LiMn 2 O 4 electrode, indicating that the SEI layer was produced during the cycling tests. Fig. 4c shows an example of the soft SEI layer in thickness of 6.5 nm. In the latter case, the AFM tip penetrated into the soft SEI with a small force, while further indentation required a large indentation force, due to the relatively hard LiMn 2 O 4 . 16 The nano-indentation experiments were repeated 100 times on the LiMn 2 O 4 surface and various types of force-distance profile were found (Fig. 5). The indentation was performed at randomly selected positions, while the distance between neighboring indenting points was maintained at approximately 120 nm. The figure indicated that the surface was often covered by multiple SEI layers with different stiffness. The topographical effect of the LiMn 2 O 4 film appears small, in this case, because the average surface roughness (R a = 4.5 nm) is smaller than the tip radius (∼10 nm), and the particle size (∼50 nm) is larger than the tip radius (Fig. 4a).
Different SEI thickness and stiffness observed in Fig. 5 indicate that the compositions of the layers are different (organic and       24 The SEI thickness, measured by the nano-indentation technique using an AFM, showed a broad distribution and its range increased with an increasing number of cycles. The histogram of SEI thickness as a function of cycle number, shown in Fig. 6, shows a broader SEI thickness distribution as the cycling tests continued, while the average thickness increased (an inset of Fig. 6), suggesting gradual SEI growth during cycling. The SEI thickness obtained in this study by indentation was thicker than the reported values measured by other techniques, such as ellipsometry and TEM, 9,11,22 which reported that the SEI thickness was in the range of 2-5 nm, much smaller than the result in this study. The relatively thin SEI from TEM measurements probably appears because of SEI decomposition by the focused electron beam under high vacuum condition, 10 while the ellipsometry technique seems to underestimate the thickness due to an assumption of an optically homogeneous and isotropic SEI.
A numerical summary of Fig. 3 and Fig. 6 is given in Table I. Very thin surface layer (∼1.3 nm) is detected from the electrode before cycle. It has been known that thin surface layer can be formed on electrode surface by contact with electrolyte. 11,12,22 The growth of the SEI during the initial cycle is rapid (1.526 nm/cycle), suggesting that the low Coulombic efficiency during the initial cycling test can be attributed to rapid SEI growth on the LiMn 2 O 4 surface. This is consistent with the studies that correlated SEI formation with Coulombic efficiency. [25][26][27] The poor Coulomb efficiency found in the early stages of the cycling test appears to be associated with the chemical reaction to produce a thin SEI, in which the ions required to run charge-discharge reactions are depleted at the interface, since they are consumed to produce the SEI.
The SEI growth rate during the first 20 cycles was 0.47 nm/cycle, which was about twice as fast as the growth rate measured during the second 20 cycles (0.19 nm/cycle). This suggests that the change of the average SEI thickness (or its growth rate) during cycling tests is rapidly reduced after 20 th cycle, indicating that the SEI is thick enough to restrict electron tunneling across the layer. 28 Considering the significant capacity fading during early cycles and stabilization of the capacity retention after 20 th cycle (Fig. 3, Table I), these results suggest that a fully grown SEI, i.e. an SEI with sufficient thickness to diminish SEI growth rate, is required for protection of the electrode against Mn 2+ dissolution and corrosion of the cathode, indicating that, once the average SEI thickness grew sufficiently to protect side reactions, the passivated SEI is preserved.

Conclusions
The change of the SEI thickness during cycling and its relation with electrochemical properties of LiMn 2 O 4 were studied by nanoindentation experiments using an AFM. The SEI produced on the LiMn 2 O 4 surface after cycling tests showed broad distributions in terms of thickness and mechanical properties, which suggest considerable variation of the SEI composition. The SEI were often composed of double layers and the distribution of SEI thickness increased with cycling tests, which appeared to be a cause of inconsistency of SEI thickness reported by many researchers. The SEI growth was rapid during initial cycling and accompanied low Coulombic efficiency. The average thickness of the SEI increased until 20 th cycle and accompanied significant capacity fading. On the other hands, the SEI growth rate was decreased and the capacity retention was stabilized after the 20 th cycle, suggesting that the long-term stabilization of the