- Email: [email protected]
In operando measurements of kinetics of solid electrolyte interphase formation in lithium-ion batteries
Journal of Power Sources 400 (2018) 426–433
Contents lists available at ScienceDirect
Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour
In operando measurements of kinetics of solid electrolyte interphase formation in lithium-ion batteries
T
Tibebu Alemua, Sylvia Ayu Pradanawatia,b, Shih-Chang Changa, Pin-Ling Lina, Yu-Lin Kuoc, Quoc-Thai Phama, Chia-Hung Sud, Fu-Ming Wanga,e,∗ a
Graduate Institute of Applied Science and Technology, National Taiwan University of Science and Technology, Taipei, Taiwan Department of Physics Energy Engineering, Surya University, Banten, Indonesia c Department of Mechanical Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan d Graduate School of Biochemical Engineering, Ming Chi University of Technology, New-Taipei City, Taiwan e Sustainable Energy Center, National Taiwan University of Science and Technology, Taipei, Taiwan b
H I GH L IG H T S
G R A P H I C A L A B S T R A C T
operando SPR and EQCM reveal the • Inin-time kinetic reaction of electrolyte. SPR and EQCM predicts • InSEIoperando formation for graphite and Si anodes.
is a good salt additive for Si • LiMB anode to inhibit the SEI formation. is a good salt additive for gra• LiTFB phite anode to reinforce the SEI formation.
A R T I C LE I N FO
A B S T R A C T
Keywords: Surface plasma resonance Quartz crystal microbalance Solid electrolyte interphase In-operando Lithium ion battery
This study applied two in operando techniques to reveal the reaction kinetics of solid electrolyte interphase formation on electrolyte and benzimidazole salt additives. The results obtained from studying interface effects reveal changes in solid electrolyte interphase mass, reflection angle, and reflection intensity within the electrolyte additives in accordance with electron-withdrawing and electron-donating substitutions. Surface plasma resonance results reveal that the electrolyte containing the electron-withdrawing salt additive exhibited the highest rate constant (774 s−1) of the binding reaction between the benzimidazole additive and Au surface, indicating the strong reaction effects on Au. The electrolyte containing the electron-withdrawing salt additive accelerates and facilitates the dissociation reaction of the ethylene carbonate–lithium ion (EC–Li+) ionic cluster. From the quartz crystal microbalance results, the electrolyte containing the electron-withdrawing salt additive shows the greatest solid electrolyte interphase mass (14.84 μg cm−2), representing the intense dissociation reactions of the EC–Li+ ionic cluster as well as solid electrolyte interphase formation and recombination. In this study, selecting a high rate constant and high binding strength of the EC–Li+ ionic cluster on the electrode surface enhance solid electrolyte interphase formation and battery performance.
∗
Corresponding author. IB 606, 43 Keelung Road, Section 4, Taipei, 106, Taiwan. E-mail address: [email protected] (F.-M. Wang).
https://doi.org/10.1016/j.jpowsour.2018.08.039 Received 3 May 2018; Received in revised form 21 July 2018; Accepted 12 August 2018 0378-7753/ © 2018 Elsevier B.V. All rights reserved.
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Scheme 1. The chemical structures of (a) Lithium Benzimidazole (LiB), (b) lithium trifluoromethyl benzimidazole (LiTFB), (c) Lithium methyl benzimidazole (LiMB).
1. Introduction
surface. In this study, in operando SPR and QCM were adopted to determine the reaction sequences during SEI formation in a case without a salt additive (the control sample) and in that with Li BZ salt additives (electron-donating and electron-withdrawing effects). By interpreting the data regarding the changes in mass, reflection angle, and reflection intensity, we determined the individual reaction rate constant, film thickness, mass change, and detailed reaction mechanisms of SEI formation on the Au electrode surface. In addition, the battery measurements on impedance analysis and rate testing are also investigated.
A solid electrolyte interphase (SEI) is a layer created on the surface of active materials during the initial electrochemical reaction in lithium-ion batteries (LIBs). It is a key component that dominates battery performance in terms of tailored functions such as ionic diffusivity [1], electrochemical stability [2], and safety [3]. Notably, a stable SEI composed of organic materials such as R-OCO2Li protects active materials from electrolyte side reactions in anode and cathode materials [4]. In LIBs, LiPF6 is the most widely used salt, and this is due to its excellent SEI formation [5], excellent solubility in carbonate solvents [6], wide electrochemical operating window [7], and high ionic conductivity [7,8]. However, it may sustain thermal stability problems [9–11] and additional side reactions [8]. The side product PF5 (a Lewis acid) usually reacts with the SEI in the presence of moisture to form HF, POF3, CO2, and LiF [8,12]. A study reported that a benzimidazole (BZ)-based Li salt additive plays a critical role in forming a new SEI film while maintaining the chemical stability of Li salt because of a Lewis acid–base reaction [12]. The study revealed a pentafluorophosphate BZ anion synthesized through a Lewis acid–base reaction between the BZ anion and PF5. The new compound, pentafluorophosphate BZ anion, was reported to inhibit the decomposition of LiPF6 by inhibiting PF5 side reactions, leading to a well-maintained battery performance. Additionally, the electrochemical behaviors of LiPF6, lithium BZ, lithium methyl BZ (LiMB; electron-donating), and lithium trifluoro methyl BZ (electronwithdrawing) salt additives were investigated. According to the report, the fluoromethyl (-CF3) substitution facilitates the realization of a high electron cloud density on the structure to resist the electron releases from bezimidazole in oxidation (anodic) reactions and further neutralize PF5 to form a stable SEI layer. It can also cause two adjacent (C–N) bond elongations because of the steric repulsion. This repulsion effect produces a lower ion-pair dissociation in accordance with weaker coordination because of extensive charge delocalization [13,14]. The electro-donating functional group (-CH3) substitution engenders less electron negativity on the imidazole ring compared with -CF3, which is ineligible for accepting PF5. Therefore, PF5 can continuously react with other Lewis bases in the electrolyte [8]. Conventional techniques such as transmission electron microscopy (TEM) [15,16], spectroscopic ellipsometry [17], and X-ray reflectivity analysis [18] can detect SEI behavior. There are some publications describing the in situ or in operando observations for the evaluations of battery materials such as Micro-Raman [19], interferometry [20], neutron scattering [21], and synchrotron radiation with thermal imaging [22], respectively. However, an SEI can be decomposed under ultra-high vacuum conditions to a nanometer size, which makes it difficult to perform a detailed exploration [17,23]. The aforementioned techniques have been extensively applied to study the mentioned parameters, because of the emergence of convenient modern in situ evaluations and in operando analyses. However, few in situ or in operando studies have ascertained real-time kinetic information on SEI formation. To realize precise in operando measurements of the kinetics of SEI formation, we developed two in operando measurement processes, namely surface plasmon resonance (SPR) and quartz crystal microbalance (QCM). These techniques were used to investigate the interface properties of salt dissociation, ionic diffusion, SEI formation, and binding reaction mechanisms on electrodes by using data regarding changes in mass, reflection angle, and intensity on the electrode
2. Experimental section 2.1. Electrolyte preparation Four electrolytes were prepared for the in operando SPR and QCM experiments. Sample A was 0.1 M lithium hexafluorophosphate (LiPF6) in ethylene carbonate (EC) and diethyl carbonate (DEC) (1:2 v/v%, battery grade, water < 20 ppm), which was purchased from Unionward Company in Taiwan. The remaining samples were identical to Sample A, except for the addition of 0.1 wt % LiB to Sample B, 0.1 wt % LiMB to Sample C, and 0.1 wt % lithium trifluoromethyl BZ (LiTFB) to Sample D. All electrolytes were prepared in a glove box in an Ar gas atmosphere to avoid the influence of moisture. The chemical structures of all BZ additives are shown in Scheme 1. 2.2. Instrumentation A Navi™ 200 from BioNavis Ltd. was used in the current study, which is the same as the method demonstrated Wang et al. [3]. As presented in the paper, a Kretschmann configuration design is used for in operando EC-SPR measurement. A gold chip (working electrode) was installed to detect the time-resolved SPR angle shifts and intensity variations during SEI formation. The laser beam (wavelength 670 nm) was p-polarized before entering the prism. Attenuated total reflection (ATR) occurred once the light beam propagated into the prism and struck a higher refractive index within the gold. The intensity changes of the incidence angle during ATR were then monitored. A slight change at the interface (a change in the refractive index or a formation of a nanoscale film thickness) would cause a change in the SPR signal, enabling precise measurements of thin-film properties as well as surface molecular interactions in real-time. A three-electrode electrochemical cell setup for QCM measurements (QCA922, Seiko) was used. Electrochemical impedance spectroscopy (EIS) was performed through a Biologic VMP3 in a frequency range of 100 M to 0.01 Hz along with an AC amplitude of 5 mV at 25 °C. All EIS measurements were performed using a half-cell (CR2032) comprising carbon and Li metal electrodes (area = 1.0 cm2) in a 100% depth of discharge. The graphite anode consisted of 93 wt % mesocarbon microbeads ΣMCMB-2528, Osaka GasΠ, 3 wt % KS4 as a conductive additive, and 4 wt % PVDF as a binder. The Si anode consisted of 85 wt % silicon ΣAldrich, particle diameter 10–30 μmΠ, 7 wt % KS6 as a conductive additive, and 8 wt % CMC as a binder. The electrolyte was 1.1 M lithium hexafluorophoshate ΣLiPF6Π in EC:EMC Σ1:2 in volumeΠ mixed solvents. A charge-discharge test was conducted in a constant current-constant voltage mode with a voltage range of 0.005–3.000 V at 0.1C/0.1C, measured using a U-bic battery tester at room temperature. 427
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
of Sample A was calculated to be 73.8 s−1 (a first-order reaction). This result is similar to the finding of a previous [3]. However, the BZ salt additives considerably changed the status of the rate constant in accordance with the interactions of BZ and the Au surface. As indicated in the Slope I regions in Fig. 2b–d, all curve tendencies are changed to an association behavior (the reflection angle increased) as the voltage increased, indicating that the BZ additives in the electrolyte had a tendency to react with Au. Table 1 shows that Samples B, C, and D delivered association rate constants of 89.9, 85.6, and 774.0 s−1 (a firstorder reaction), respectively, revealing two pieces of vital information regarding the effects of BZ additives. We assumed that a competition reaction induced the BZ additives to diffuse close to the Au surface owing to the electronegativity of the imidazole ring of the additives and adsorb first on the surface before EC molecules. After the binding reaction of the BZ additives on the Au surface was completed, the EC–Li+ ionic clusters were attracted by the imidazole structure of the BZ additives owing to the Lewis base structure and the repulsion of the dissociation reaction. The repulsion time was approximately 4–10 min. The samples can be ordered as follows in terms of repulsion time: Sample D < Sample B < Sample C. This thus indicates that Sample D had the shortest repulsion time. In addition, with different functional group designs on the BZ additives, the association rate constant demonstrated that the high electro-withdrawing group (-CF3, Sample D) dramatically enhanced the rate of the binding reaction between BZ and the Au surface. Compared with Sample B (-H), the association rate of Sample D increased by 8–9 times. Conversely, the electro-donating group design (-CH3, Sample C) extended the binding reaction time owing to the strong attractive force with Au. As indicated in the Slope II region, Sample A showed a continuing dissociation reaction (reflection angle decreased) of the EC–Li+ cluster corresponding to a decreasing potential. In this region, the electrochemical voltage triggered the decomposition of the EC and formed an SEI on Au. Therefore, the reflection angle change observed for Slope II was lower than that observed for Slope I; this can be attributed to the EC decomposition effect. From the results presented in Table 1, the dissociation rate constant of the EC–Li+ cluster in this reaction range decreased to 32.8 s−1. According to the literature [5,31,32], an SEI consists of several organic and inorganic lithium carbonates from its two-step reactions, indicating that SEI formation drastically influences the dissociation process of EC–Li+ clusters. Notably, three BZ additives displayed individual behaviors in the Slope II region, indicating that the functional group substitutions played critical roles in the electrochemical reaction. With a strong electro-withdrawing group (-CF3) effect in Sample D, the spectrum demonstrated a dissociation reaction process; -CF3 was effectively used to repulse the lithium ion, which dissociated from its original EC–Li+ cluster. Previous research identified, through a simulation, that LiTFB has the lowest dissociation energy and the longest bond length of Li anions compared with other BZ additives [8], implying the lithium ions are easily released from the BZ structure. The dissociated EC was then observed to be decomposed and react with single Li+ to form the SEI after this dissociation process. The dissociation rate constant of Sample D in the Slope II region was calculated to be 48.8 ng−1s−1 (a second-order reaction). Compared with the Slope I region, the reflection angle observed for the Slope II tended to change dramatically (Δθ = 0.063°), indicating that -CF3 effectively inhibited the solvation and association of LiTFB and the EC–Li+ cluster and triggered SEI formation. However, Sample B in the Slope II region demonstrated a similar result to Sample D: the rate constant changed from an association reaction to a dissociation reaction and the reflection angle decreased (a second-order reaction). Nevertheless, without the -CF3 substitution, the electron cloud density of LiBZ was not strong enough to continuously repulse the lithium ions from its original EC–Li+ cluster, owing to the electron hole on the proton (-H) position of the BZ structure. Therefore, the reflection angle change observed for this Slope II dissociation reaction tended to be weaker (Δθ = 0.023°) and lasted a shorter time. This result confirms that SEI formation was
2.3. In operando cells Polyether ether kethone constituted the in operando SPR and QCM cells, which were insulated with an NBR-70/Kalrez type O-ring. A Pt pseudoreference (Pt pseudo) electrode and Pt counter electrode (CE) [24] were used in this study. The Au working electrode, Pt CE, and Pt pseudo electrode were directly connected to a multichannel potentiostat (Biologic VMP3). Before the experiment was initiated, the control sample and modified electrolyte mixtures were injected into the cell (bottom left corner) until the space was filled. Notably, Kalrez O-rings are the best option for making seals because they are resistant to swelling in the presence of organic electrolytes. In this study, we used a Pt electrode as a pseudoreference to compare all the recorded potentials with the Ptpseudo potential. According to a report and an investigation [3,25,26], an EC undergoes a reduction reaction at −1.24 V (equal to 0.76 V vs Li/Li+). For the safety of the EC-SPR compartments, we used diluted 0.1 M LiPF6 in EC:EMC (1:2 in volume) electrolytes for Sample A and 0.1 wt % of BZ salt additive for Samples B, C, and D. 2.4. Electrochemical operations and data analysis In operando QCM measurements were performed using the frequency change of the linear sweep voltammetry (LSV) process (scanning rate, 1 mV s−1; voltage range, 0 to −2 V). Through the Sauerbrey equation, the frequency changes derived from the experimental results were converted into mass changes [27]. We adopted 9 MHz of Au chips, and the theoretical mass sensitivity (ks) was considered to be 18.31 × 107 Hz g−1 cm−2. Δf = -ks × Δm (Δf, frequency changes; ks, theoretical mass sensitivity; and Δm, mass changes). In operando SPR measurements were performed in the angular scanning mode to evaluate angle (θ) changes during the LSV process (scanning rate, 1 mV s−1; voltage range, 0 to −2 V). To analyze the refractive index, thin-film thickness, and angular scanning curves, Winspall software (ver. 3.02) was used to simulate the SPR curves required for calculating the mass and thickness changes. Winspall software is a simulation package that facilitates calculations regarding SPR performance based on Fresnel's equation. This software computes the reflectivity of optical multilayer systems. The thickness and refractive index of the Au layer was determined using a Bionavis™, which was used as a reference for our simulation. Basic kinetic information on the binding rate constant owing to angle changes could also be obtained from sensogram data with the help of TraceDrawer (ver. 1.8.2. X) software. It helped us evaluate, summarize, and report the rate constant and resonance angle changes from the real-time experimental outcomes of physical and chemical interactions at the interface. 3. Results and discussion 3.1. Analysis of in operando SPR Fig. 1 shows the in operando sensogram profiles of all the electrolyte systems in the SPR measurements. After the electrochemical reaction was triggered, an independent dissociation process was observed for sample A (the reflection angle decreased), as indicated by the Slope I region (0–9.5 min) in Fig. 1a. At the stage of electrolyte preparation, LiPF6 underwent dissociation, and the dissociated single lithium ions were solvated by the EC. EC–Li+ ionic clusters were formed due to the high dielectric constant of the EC [3,28–30] with Li+ and diffused close to the Au electrode surface because of the concentration gradient effect. The EC–Li+ ionic clusters were then dissociated to single lithium ions by the electrochemical reaction on the electrode surface. The dissociation rate constant of the EC–Li+ ionic cluster in the Slope I region 428
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Fig. 1. In-operando sensogram profiles of reflection angle of samples (a) A, (b) B, (c) C, and (d) D using 1 mVs-1 as the scanning rate in the electrochemical window of 0 - (−2)V.
15.8 and 16.8 ng−1s−1, respectively. We can conclude that the Li-Au alloying reaction was independent of the effects of LiTFB and LiBZ salt additives. However, Sample C still illustrated a continuous association reaction, indicating that the LiMB salt additive exhibited considerable performance in scavenging the EC–Li+ cluster. In this system, we expected that no more Li-Au alloying layer would be formed. The association rate constant of Sample C in the Slope III region was calculated to be 61.8 s−1. Table 2 shows the simulation results for the forming layer structure on the Au electrode surface after the electrochemical reaction on BZ addition. The thicknesses of the adhesion material, Au, SEI, and Li-Au alloying layers were estimated at 1.53, 49.61, 8.6, and 0.03 nm, respectively, in Sample A. In Samples B, C, and D, the SEI thicknesses were 2.73, 2.60, and 5.20 nm, respectively, which were determined to be consistent with the results shown in Fig. 1d that LiTFB in Sample D had the greatest performance in SEI formation. In addition, the thickness of the Li-Au alloying layer in Sample C was nearly 0 nm, validating the results in Fig. 1c.
triggered in the Slope II region of Sample B, but the reaction level was lower than that of Sample D. The dissociation rate constant of Sample B in the Slope II region was calculated to be 25.6 ng−1s−1 (a second-order reaction). Furthermore, with a strong electron-donating group (-CH3) effect, the Slope II region of Sample C displayed a continuous association reaction, indicating that -CH3 was used to strongly scavenge the lithium ions from the original EC–Li+ cluster. Previous research also identified, through a simulation, LiMB to have the largest dissociation energy and the shortest bond length of Li anions compared with other BZ additives [8], implying that lithium ions are barely released from the BZ structure. We expected that the two integral association reactions in the Sample C system would show the EC is decomposed shortly, revealing that the strong -CH3 substitution effectively inhibited the desolvation and further decomposition of the EC. The association rate constant of Sample C in the Slope II region was calculated to be 70.3 s−1 (a first-order reaction). In the Slope III region, Sample A showed a continuous dissociation reaction, which is consistent with the previous Slope I and II regions, revealing that the lithium ion was dissociated from its original EC–Li+ cluster and joined the Li-Au alloying reaction below −2 V (vs Pt/Li+). In this reaction step, the dissociation rate constant was calculated to be 17.0 s−1, which was much lower than the values in the Slope I and II regions. We supposed that the Li-Au alloying layer formed on the electrode surface when the potential was approached to −2 V. The LiAu alloying reaction disturbed the dissociation of the EC–Li+ cluster and decreased the reaction rate. Samples B and D demonstrated similar results in the Slope III region, in which the rate constant values were
3.2. Analysis of in operando QCM Fig. 2 shows the in operando QCM measurements for Samples A, B, C, and D at voltages ranging from 0 to −2 V and a scanning rate of 1 mVs−1 on the Au working electrode. Fig. 2a presents two reduction reactions: SEI formation at −0.81 V and Li-Au alloying reaction at −1.93 V. Apart from these two reactions, two resting (zero current 429
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Fig. 2. In-operando mass profiles of the samples (a) A, (b) B, (c) C, and (d) D using 1 mVs-1 as the scanning rate in the electrochemical window of 0 - (−2)V.
flows) regions were observed according to the LSV curve. From the mass curve, the reactions could be divided into four regions in terms of the change in the mass curves. The reaction in the first region (0–23 min; 0 to −0.85 V) produced a mass of 1.31 μg cm−2 with zero current behavior, indicating that no electron transfer occurred in this voltage area, which can be correlated with the EC–Li+ cluster adsorption on the Au surface. In the reaction in the second region (23–26 min; −0.85 to −1.13 V), the mass curve showed a sudden increase engendered by the EC decomposition. In this region, the SEI formed a mass of 1.52 μg cm−2 with a nonzero current flow of electrochemical reaction. As the voltage increased, the mass curve continuously increased and tended to approach the second zero current area, which is the third region (26–41 min; −1.13 to −1.62 V). In this region, the reaction was neither an electrochemical reaction nor an EC–Li+ cluster adsorption. This was defined as SEI recombination, because SEI restacking or reconstruction occurred and engendered an increase in density. In this region, SEI recombination generated a mass of 1.31 μg cm−2. When the potential approached −2 V, the reaction in the fourth region (41–50 min; −1.62 to −2 V) involved rapid and highintensity Li-Au alloying formation, indicating the lithium ions dissociated from the EC–Li+ cluster and joined the reaction while the
Table 1 The kinetic parameters measured from in-operando SPR of all samples. Slope
1 2 3
Samples A
B
C
D
Rate constant
Rate constant
Rate constant
Rate constant
73.8 32.8 17.0
89.9 25.6 15.8
85.6 70.3 61.8
774.0 48.8 16.8
Hint: The units of 1st order dissociation/association rate constant (s-1) and 2nd order dissociation rate constant (ng-1 s-1). Table 2 Winspall simulation results for samples A, B, C, and D.
Adhesion layer/nm Au/nm SEI/nm Au-Li alloying/nm
A
B
C
D
1.53 49.61 8.60 0.03
1.53 49.75 2.73 0.03
1.53 50.01 2.60 ∼0
1.53 49.83 5.20 0.04
Table 3 The kinetic parameters measured from in-operando QCM of all samples. Reaction Sample
EC-Li + cluster/μg
EC decomposition/μg
SEI recombination/μg
Li-Au alloying/μg
Total/μg
A B C D
1.31 0.64 0.44 1.42
1.52 0.49 0.43 1.05
1.31 0.80 0.30 2.45
6.56 2.20 2.02 9.92
10.70 3.89 3.43 14.84
430
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Fig. 3. Contact angle measurements of samples (a) A, (b) B, (c) C, and (d) D on Au chip with electrolyte.
potential reached exactly 0 V (Li/Li+). In this region, the Li-Au alloying formation reaction produced a mass of 6.56 μg cm−2. According to these results, Sample A produced a total mass of 10.70 μg cm−2, as shown in Table 3. Regarding the effect of the BZ additives, the in operando QCM measurements showed noteworthy results for the various samples compared with the Sample A system. In Sample D (Fig. 2d), LiTFB was strongly adsorbed on the Au surface owing to its imidazole ring and -CF3 electron-withdrawing group electronegativity properties. After the LiTFB binding reaction on the Au surface was completed, the EC–Li+ ionic clusters were easily attracted by the imidazole structure owing to its Lewis base structure and repulsion of dissociation in a short time range. Therefore, the mass in the first region (0–22 min; 0 to −0.81 V) was calculated to be 1.42 μg cm−2, indicating a greater mass and longer reaction time for EC–Li+ cluster adsorption compared with Sample A. Sample D thus showed the fastest association rate in this reaction region, as shown in Table 1. These results indicate that LiTFB provided a strong interaction with the EC molecule and delayed the dissociation reaction of lithium ions in this zero-current reaction step. As shown in Fig. 2b and c, compared with Sample D, the electron-donating effects in Samples C and B resulted in extremely low mass values for EC–Li+ cluster adsorption on Au, with the values being 0.44 and 0.64 μg cm−2, respectively. This result is consistent with the in operando SPR measurement results that the electron-donating substitutions were not used to enhance lithium ion dissociation because the LiMB salt additive increased the association reaction with the EC–Li+ cluster and Au surface. In the reaction of the second region, with the current effects, the LSV curve obtained for Sample D showed late EC decomposition at −1.14 V with a strong reaction intensity. In addition, EC decomposition resulted in a mass of 1.05 μg cm−2 on Au at 22–29 min and −0.81 to −1.18 V from the mass curve, as shown in Table 3. In this reaction, -CF3 effectively inhibited the solvation and association of imidazole structures and Li+ but triggered EC decomposition. The EC in Sample D showed a strong electrochemical reaction, with the mass curve exhibiting an increasing tendency, as shown Fig. 2d. From the SPR result, the dissociation rate constant of Sample D was the highest. Conversely, Sample C (-CH3) and Sample B (-H) had EC decomposition mass values of 0.43 and 0.49 μg cm−2 on Au, as shown in Fig. 2b and c. As the electron-donating effect of the electrolyte increased, the intensity of EC decomposition decreased, especially in Sample C with a whole and complete association process for the electrochemical reaction, as shown in Fig. 1c. In the reaction in the third region, without the current effects, the mass curve derived for Sample D showed a continuous increase, indicating that the EC decomposition products may had a loose structure and that restacked or recombined to form a new SEI structure on Au. In this reaction, SEI recombination generated a mass of 2.45 μg cm−2 at 29–46 min and −1.18 to −1.83 V. Notably, Samples B and C displayed less mass in SEI recombination for the same reason as the reaction in the second region. The mass values engendered by the SEI recombination reaction were 0.8 and 0.3 μg cm−2 in Samples B and C, respectively. When the potential approached −2 V, the reaction in the fourth region (46–50 min; −1.83 to −2 V) in Sample D involved a more rapid
and higher-intensity Li-Au alloying formation process compared with Sample A, indicating that the -CF3 substitution greatly enhanced the lithium ion dissociation from the EC–Li+ cluster and enabled the ions to join the reaction, when the potential reached exactly 0 V (Li/Li+). In this reaction, the Li-Au alloying formation in Sample D resulted in a mass of 9.92 μg cm−2. According to these results, Sample D produced a total mass of 14.84 μg cm−2 (Table 3). In Samples B and C, the mass values engendered by Li-Au alloying formation were 2.20 and 2.02 μg cm−2, respectively. According to these results, Samples B and C produced a total mass of 3.43 and 3.89 μg cm−2 (Table 3). 3.3. Analysis of contact angle measurement Fig. 3 shows the contact angles of Samples A, B, C, and D with the Au electrode surface. The contact angle measurement was conducted to test electrolyte compatibility with the Au surface and confirm the in operando SPR results. As presented in Fig. 3c, Sample C showed the lowest contact angle (14.0°) with the Au surface, followed by Samples B (31.4°) and D (32.3°). This result implies that Sample C delivered high binding energy with Au, indicating the associated energy was also high. From the in operando SPR results in Fig. 1c, the sensorgram profile showed a continuous association reaction with the electron-donating substitution effect on the additive structure, and the contact angle measurement confirmed this behavior. Compared with Sample D, the electron-withdrawing substitution considerably raised the contact angle that represented the use of LiTFB, which enhanced the repulsion of lithium ions from the EC–Li+ cluster on the Au. 3.4. Surface energy of Au and graphite To confirm the in operando SPR and QCM measurement results, identifying the surface energy of Au and graphite was crucial. A previous study reported that Au has a surface energy of 1610–1790 mN m−1, whereas graphite has an energy of 50.8 mN m−1 [33–35]. As presented in Fig. 3, Sample C demonstrated the lowest contact angle with the Au surface, revealing that the surface energy of the electrolyte was increased by adding LiMB. Conversely, the LiTFB additive with fluorine substitution dramatically reduced the surface energy, which showed high contact angle behavior. Hsieh et al. has revealed that -CF3 group substitution considerably decreases the surface tension due to the F electronegativity property [36]. However, graphite has a lower surface energy than Au, indicating the high surface energy of Sample C renders it adequate for binding with Au but not with graphite. In this study, Sample D containing electron-withdrawing substitution effects effectively decreased the surface energy and was determined to be suitable for binding with graphite. From this conclusion, the contact impedance of Sample D was expected to be lower than that of Sample C in the graphite system. 3.5. Battery performances in MCMB and Si anode half-cells Fig. 4a shows the impedance analysis of MCMB anode half-cells after the first cycle test. According to the results, Sample D showed the 431
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Fig. 4. MCMB half cell electrochemical measurements of (a) the impedance analysis/simulation fitting results after the first cycle and (b) the cycle performance at room temperature.
lowest SEI (Rsei) resistance (1.85 Ω), as determined from an equivalent circuit model simulation, revealing the chemical and physical properties of the formed SEI were notably reinforced by the LiTFB enhancements, such as ionic diffusivity. In addition, the charge transfer resistance (Rct) of the electrochemical reaction demonstrated that the solvation and association between the imidazole structure and EC–Li+ cluster on the MCMB surface could be eliminated and triggered the reaction of the de-solvated EC and lithium ions. However, the electrondonating group substitution considerably affected the impedances of Rsei and Rct on the Sample C system, indicating that in operando SPR and QCM are suitable for predicting battery performance. Fig. 4b shows the cycle performance of all modified battery systems. The results demonstrate that Sample D involving the electron-withdrawing substitution effect provided excellent dissociation of the EC–Li+ cluster, SEI formation, and contact ability with graphite. However, Si plays different rule in dominating the battery performance compares with the graphite. In accordance with the literature [37,38], Si suffers tremendous volume expansion during the electrochemical reaction. The electrolyte contacts new surface and forms SEI continuously. In this study, Sample C may use to inhibit the dissociation of the EC–Li+ cluster in the electrolyte and form less SEI on Si surface. Fig. 5a shows the first charge and discharge curves that the Sample C is able to enhance the coulombic efficiency to 84.8% and achieve the capacity to 2772.9 mAh g−1. In the meantime, the Samples A and D only demonstrate the capacities to 2407.7 and 2722.2 mAh g−1, respectively. With
cycle test, Fig. 5b indicates that the Sample C displays better cycle performance owing to the elimination of SEI on Si surface. 4. Conclusion This is the first study to apply in operando SPR and QCM to investigate electrolyte additive adsorption and solvation processes on Au, EC–Li+ dissociation reactions on Au, the kinetics of EC–Li+ dissociation reactions, SEI formation and Au-Li alloying reactions, and rate constants. Sample A had the thickest SEI layer, followed by Samples D, B, and C. Sample C possessed high surface binding energy as a result of an electron-donating substitution effect within the electrolyte on the Au surface, as well as its high association reaction with the EC–Li+ cluster. Sample D showed considerable Li+ free movement toward and away from the Au surface because of its high reaction constant. Therefore, an electrolyte containing electron-withdrawing effects could potentially improve lithium ion battery performance with graphite anode. Interestingly, Sample C showed the reverse behavior, it is suitable for Si anode in eliminating the SEI formation. In conclusion, the investigations of the stated parameters using in operando SPR and QCM were successfully accomplished under a low electrolyte concentration as well as a tiny amount of additive. These novel measurements processes provide extreme sensitivity, which could not be achieved with conventional electrochemical techniques.
Fig. 5. (a) The 1st cycle and (b) cycle performance of Si anode half cells. 432
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Acknowledgment
[17] J. Lei, L. Li, R. Kostecki, R. Muller, F. McLarnon, J. Electrochem. Soc. 152 (2005) A774–A777. [18] M. Hirayama, N. Sonoyama, M. Ito, M. Minoura, D. Mori, A. Yamada, K. Tamura, J.i. Mizuki, R. Kanno, J. Electrochem. Soc. 154 (2007) A1065–A1072. [19] Y. Miroshnikov, J. Yang, V. Shokhen, M. Alesker, G. Gershinsky, A. Kraytsberg, Y. Ein-Eli, D. Zitoun, ACS Appiled Energy Materials 1 (2018) 1096–1105. [20] H. Jung, K. Gerasopoulos, X. Zhang, R. Ghodssi, J. Phys. Conf. 660 (2015) 012065. [21] A. Senyshyn, M.J. Muhlbauer, K. Nikolowski, T. Pirling, H. Ehrenberg, J. Power Sources 203 (2012) 126–129. [22] D.P. Finegan, M. Scheel, J.B. Robinson, B. Tjaden, I. Hunt, T.J. Mason, J. Millichamp, M.D. Michiel, G.J. Offer, G. Hinds, D.J.L. Brett, P.R. Shearing, Nature Commmunications 6 (2015) 6924. [23] M. Dollé, S. Grugeon, B. Beaudoin, L. Dupont, J.M. Tarascon, J. Power Sources 97–98 (2001) 104–106. [24] R.L. Sacci, J.M. Black, N. Balke, N.J. Dudney, K.L. More, R.R. Unocic, Nano Lett. 15 (2015) 2011–2018. [25] A.M. Haregewoin, E.G. Leggesse, J.C. Jiang, F.M. Wang, B.J. Hwang, S.D. Lin, Electrochim. Acta 136 (2014) 274–285. [26] P. Bach, M. Stratmann, I. Valencia-Jaime, A.H. Romero, F.U. Renner, Electrochim. Acta 164 (2015) 81–89. [27] F.M. Wang, D.T. Shieh, J.H. Cheng, C.R. Yang, Solid State Ionics 180 (2010) 1660–1666. [28] W. Cui, Y. Lansac, H. Lee, S.T. Hong, Y.H. Jang, Phys. Chem. Chem. Phys. 18 (2016) 23607–23612. [29] R. Naejus, R. Coudert, P. Willmann, D. Lemordant, Electrochim. Acta 43 (1998) 275–284. [30] N. Kumar, J.M. Seminario, J. Phys. Chem. C 120 (2016) 16322–16332. [31] F.M. Wang, H.M. Cheng, H.C. Wu, S.Y. Chu, C.S. Cheng, C.R. Yang, Electrochim. Acta 54 (2009) 3344–3351. [32] C.-S. Cheng, F.-M. Wang, J. Rick, Int. J. Electrochem. Sci 7 (2012) 8676–8687. [33] A. Kozbial, Z. Li, C. Conaway, R. McGinley, S. Dhingra, V. Vahdat, F. Zhou, B. D'Urso, H. Liu, L. Li, Langmuir: The ACS Journal of Surfaces and Colloids 30 (2014) 8598–8606. [34] A. Kozbial, F. Zhou, Z. Li, H. Liu, L. Li, Accounts Chem. Res. 49 (2016) 2765–2773. [35] J. Lee, B. Lee, Carbon letters 21 (2017) 107–110. [36] C.T. Hsieh, J.M. Chen, R.R. Kuo, T.S. Lin, C.F. Wu, Appl. Surf. Sci. 240 (2005) 318–326. [37] S.J. Park, H. Zhao, G. Ai, C. Wang, X. Song, N. Yuca, V.S. Battaglia, W. Yang, G. Liu, Journal of the American Society 137 (2015) 2565–2571. [38] K. Schroder, J. Alvardo, T.A. Yersak, J. Li, N. Dudney, L.J. Webb, Y.S. Meng, K.J. Stevenson, Chem. Mater. 27 (2015) 5531–5542.
The author is grateful for the financial support from the Ministry of Science and Technology (MOST) of Taiwan, R.O.C., under grant numbers 102-2221-E-011-016-MY3, 104-3113-E-011-002, 104-2745-8-011001, 105-3113-E-011-002, 105-2628-E-011-005-MY3, 105-2811-E-011017, 106-3113-E-011-001, 106-2923-E-036-002-MY3, and 106-2923-E007-005. References [1] H. Yildirim, A. Kinaci, M.K.Y. Chan, J.P. Greeley, ACS Appl. Mater. Interfaces 7 (2015) 18985–18996. [2] Q. Lv, Y. Liu, T. Ma, W. Zhu, X. Qiu, ACS Appl. Mater. Interfaces 7 (2015) 23501–23506. [3] S.A. Pradanawati, F.-M. Wang, C.-H. Su, J. Power Sources 330 (2016) 127–131. [4] P. Verma, P. Maire, P. Novák, Electrochim. Acta 55 (2010) 6332–6341. [5] S.S. Zhang, J. Power Sources 162 (2006) 1379–1394. [6] M. Xu, L. Zhou, Y. Dong, Y. Chen, J. Demeaux, A.D. MacIntosh, A. Garsuch, B.L. Lucht, Energy Environ. Sci. 9 (2016) 1308–1319. [7] R. Younesi, G.M. Veith, P. Johansson, K. Edström, T. Vegge, Energy Environ. Sci. 8 (2015) 1905–1922. [8] F.M. Wang, S.A. Pradanawati, N.H. Yeh, S.C. Chang, Y.T. Yang, S.H. Huang, P.L. Lin, J.F. Lee, H.S. Sheu, M.L. Lu, C.K. Chang, A. Ramar, C.-H. Su, Chem. Mater. 29 (2017) 5537–5549. [9] X.G. Teng, F.Q. Li, P.H. Ma, Q.D. Ren, S.Y. Li, Thermochim. Acta 436 (2005) 30–34. [10] X. Zhang, P.N. Ross, R. Kostecki, F. Kong, S. Sloop, J.B. Kerr, K. Striebel, E.J. Cairns, F. McLarnon, J. Electrochem. Soc. 148 (2001) A463–A470. [11] L. Niedzicki, P. Oledzki, A. Bitner, M. Bukowska, P. Szczecinski, J. Power Sources 306 (2016) 573–577. [12] S.A. Pradanawati, F.-M. Wang, J. Rick, Electrochim. Acta 135 (2014) 388–395. [13] L. Niedzicki, G.Z. Żukowska, M. Bukowska, P. Szczeciński, S. Grugeon, S. Laruelle, M. Armand, S. Panero, B. Scrosati, M. Marcinek, W. Wieczorek, Electrochim. Acta 55 (2010) 1450–1454. [14] X. Hu, J. Peng, Y. Huang, D. Yin, J. Liu, J. Separ. Sci. 32 (2009) 4126–4132. [15] K. Zhong, X. Xia, B. Zhang, H. Li, Z. Wang, L. Chen, J. Power Sources 195 (2010) 3300–3308. [16] J. Xu, Y. Hu, T. liu, X. Wu, Nanomater. Energy 5 (2014) 67–73.
433
Contents lists available at ScienceDirect
Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour
In operando measurements of kinetics of solid electrolyte interphase formation in lithium-ion batteries
T
Tibebu Alemua, Sylvia Ayu Pradanawatia,b, Shih-Chang Changa, Pin-Ling Lina, Yu-Lin Kuoc, Quoc-Thai Phama, Chia-Hung Sud, Fu-Ming Wanga,e,∗ a
Graduate Institute of Applied Science and Technology, National Taiwan University of Science and Technology, Taipei, Taiwan Department of Physics Energy Engineering, Surya University, Banten, Indonesia c Department of Mechanical Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan d Graduate School of Biochemical Engineering, Ming Chi University of Technology, New-Taipei City, Taiwan e Sustainable Energy Center, National Taiwan University of Science and Technology, Taipei, Taiwan b
H I GH L IG H T S
G R A P H I C A L A B S T R A C T
operando SPR and EQCM reveal the • Inin-time kinetic reaction of electrolyte. SPR and EQCM predicts • InSEIoperando formation for graphite and Si anodes.
is a good salt additive for Si • LiMB anode to inhibit the SEI formation. is a good salt additive for gra• LiTFB phite anode to reinforce the SEI formation.
A R T I C LE I N FO
A B S T R A C T
Keywords: Surface plasma resonance Quartz crystal microbalance Solid electrolyte interphase In-operando Lithium ion battery
This study applied two in operando techniques to reveal the reaction kinetics of solid electrolyte interphase formation on electrolyte and benzimidazole salt additives. The results obtained from studying interface effects reveal changes in solid electrolyte interphase mass, reflection angle, and reflection intensity within the electrolyte additives in accordance with electron-withdrawing and electron-donating substitutions. Surface plasma resonance results reveal that the electrolyte containing the electron-withdrawing salt additive exhibited the highest rate constant (774 s−1) of the binding reaction between the benzimidazole additive and Au surface, indicating the strong reaction effects on Au. The electrolyte containing the electron-withdrawing salt additive accelerates and facilitates the dissociation reaction of the ethylene carbonate–lithium ion (EC–Li+) ionic cluster. From the quartz crystal microbalance results, the electrolyte containing the electron-withdrawing salt additive shows the greatest solid electrolyte interphase mass (14.84 μg cm−2), representing the intense dissociation reactions of the EC–Li+ ionic cluster as well as solid electrolyte interphase formation and recombination. In this study, selecting a high rate constant and high binding strength of the EC–Li+ ionic cluster on the electrode surface enhance solid electrolyte interphase formation and battery performance.
∗
Corresponding author. IB 606, 43 Keelung Road, Section 4, Taipei, 106, Taiwan. E-mail address: [email protected] (F.-M. Wang).
https://doi.org/10.1016/j.jpowsour.2018.08.039 Received 3 May 2018; Received in revised form 21 July 2018; Accepted 12 August 2018 0378-7753/ © 2018 Elsevier B.V. All rights reserved.
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Scheme 1. The chemical structures of (a) Lithium Benzimidazole (LiB), (b) lithium trifluoromethyl benzimidazole (LiTFB), (c) Lithium methyl benzimidazole (LiMB).
1. Introduction
surface. In this study, in operando SPR and QCM were adopted to determine the reaction sequences during SEI formation in a case without a salt additive (the control sample) and in that with Li BZ salt additives (electron-donating and electron-withdrawing effects). By interpreting the data regarding the changes in mass, reflection angle, and reflection intensity, we determined the individual reaction rate constant, film thickness, mass change, and detailed reaction mechanisms of SEI formation on the Au electrode surface. In addition, the battery measurements on impedance analysis and rate testing are also investigated.
A solid electrolyte interphase (SEI) is a layer created on the surface of active materials during the initial electrochemical reaction in lithium-ion batteries (LIBs). It is a key component that dominates battery performance in terms of tailored functions such as ionic diffusivity [1], electrochemical stability [2], and safety [3]. Notably, a stable SEI composed of organic materials such as R-OCO2Li protects active materials from electrolyte side reactions in anode and cathode materials [4]. In LIBs, LiPF6 is the most widely used salt, and this is due to its excellent SEI formation [5], excellent solubility in carbonate solvents [6], wide electrochemical operating window [7], and high ionic conductivity [7,8]. However, it may sustain thermal stability problems [9–11] and additional side reactions [8]. The side product PF5 (a Lewis acid) usually reacts with the SEI in the presence of moisture to form HF, POF3, CO2, and LiF [8,12]. A study reported that a benzimidazole (BZ)-based Li salt additive plays a critical role in forming a new SEI film while maintaining the chemical stability of Li salt because of a Lewis acid–base reaction [12]. The study revealed a pentafluorophosphate BZ anion synthesized through a Lewis acid–base reaction between the BZ anion and PF5. The new compound, pentafluorophosphate BZ anion, was reported to inhibit the decomposition of LiPF6 by inhibiting PF5 side reactions, leading to a well-maintained battery performance. Additionally, the electrochemical behaviors of LiPF6, lithium BZ, lithium methyl BZ (LiMB; electron-donating), and lithium trifluoro methyl BZ (electronwithdrawing) salt additives were investigated. According to the report, the fluoromethyl (-CF3) substitution facilitates the realization of a high electron cloud density on the structure to resist the electron releases from bezimidazole in oxidation (anodic) reactions and further neutralize PF5 to form a stable SEI layer. It can also cause two adjacent (C–N) bond elongations because of the steric repulsion. This repulsion effect produces a lower ion-pair dissociation in accordance with weaker coordination because of extensive charge delocalization [13,14]. The electro-donating functional group (-CH3) substitution engenders less electron negativity on the imidazole ring compared with -CF3, which is ineligible for accepting PF5. Therefore, PF5 can continuously react with other Lewis bases in the electrolyte [8]. Conventional techniques such as transmission electron microscopy (TEM) [15,16], spectroscopic ellipsometry [17], and X-ray reflectivity analysis [18] can detect SEI behavior. There are some publications describing the in situ or in operando observations for the evaluations of battery materials such as Micro-Raman [19], interferometry [20], neutron scattering [21], and synchrotron radiation with thermal imaging [22], respectively. However, an SEI can be decomposed under ultra-high vacuum conditions to a nanometer size, which makes it difficult to perform a detailed exploration [17,23]. The aforementioned techniques have been extensively applied to study the mentioned parameters, because of the emergence of convenient modern in situ evaluations and in operando analyses. However, few in situ or in operando studies have ascertained real-time kinetic information on SEI formation. To realize precise in operando measurements of the kinetics of SEI formation, we developed two in operando measurement processes, namely surface plasmon resonance (SPR) and quartz crystal microbalance (QCM). These techniques were used to investigate the interface properties of salt dissociation, ionic diffusion, SEI formation, and binding reaction mechanisms on electrodes by using data regarding changes in mass, reflection angle, and intensity on the electrode
2. Experimental section 2.1. Electrolyte preparation Four electrolytes were prepared for the in operando SPR and QCM experiments. Sample A was 0.1 M lithium hexafluorophosphate (LiPF6) in ethylene carbonate (EC) and diethyl carbonate (DEC) (1:2 v/v%, battery grade, water < 20 ppm), which was purchased from Unionward Company in Taiwan. The remaining samples were identical to Sample A, except for the addition of 0.1 wt % LiB to Sample B, 0.1 wt % LiMB to Sample C, and 0.1 wt % lithium trifluoromethyl BZ (LiTFB) to Sample D. All electrolytes were prepared in a glove box in an Ar gas atmosphere to avoid the influence of moisture. The chemical structures of all BZ additives are shown in Scheme 1. 2.2. Instrumentation A Navi™ 200 from BioNavis Ltd. was used in the current study, which is the same as the method demonstrated Wang et al. [3]. As presented in the paper, a Kretschmann configuration design is used for in operando EC-SPR measurement. A gold chip (working electrode) was installed to detect the time-resolved SPR angle shifts and intensity variations during SEI formation. The laser beam (wavelength 670 nm) was p-polarized before entering the prism. Attenuated total reflection (ATR) occurred once the light beam propagated into the prism and struck a higher refractive index within the gold. The intensity changes of the incidence angle during ATR were then monitored. A slight change at the interface (a change in the refractive index or a formation of a nanoscale film thickness) would cause a change in the SPR signal, enabling precise measurements of thin-film properties as well as surface molecular interactions in real-time. A three-electrode electrochemical cell setup for QCM measurements (QCA922, Seiko) was used. Electrochemical impedance spectroscopy (EIS) was performed through a Biologic VMP3 in a frequency range of 100 M to 0.01 Hz along with an AC amplitude of 5 mV at 25 °C. All EIS measurements were performed using a half-cell (CR2032) comprising carbon and Li metal electrodes (area = 1.0 cm2) in a 100% depth of discharge. The graphite anode consisted of 93 wt % mesocarbon microbeads ΣMCMB-2528, Osaka GasΠ, 3 wt % KS4 as a conductive additive, and 4 wt % PVDF as a binder. The Si anode consisted of 85 wt % silicon ΣAldrich, particle diameter 10–30 μmΠ, 7 wt % KS6 as a conductive additive, and 8 wt % CMC as a binder. The electrolyte was 1.1 M lithium hexafluorophoshate ΣLiPF6Π in EC:EMC Σ1:2 in volumeΠ mixed solvents. A charge-discharge test was conducted in a constant current-constant voltage mode with a voltage range of 0.005–3.000 V at 0.1C/0.1C, measured using a U-bic battery tester at room temperature. 427
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
of Sample A was calculated to be 73.8 s−1 (a first-order reaction). This result is similar to the finding of a previous [3]. However, the BZ salt additives considerably changed the status of the rate constant in accordance with the interactions of BZ and the Au surface. As indicated in the Slope I regions in Fig. 2b–d, all curve tendencies are changed to an association behavior (the reflection angle increased) as the voltage increased, indicating that the BZ additives in the electrolyte had a tendency to react with Au. Table 1 shows that Samples B, C, and D delivered association rate constants of 89.9, 85.6, and 774.0 s−1 (a firstorder reaction), respectively, revealing two pieces of vital information regarding the effects of BZ additives. We assumed that a competition reaction induced the BZ additives to diffuse close to the Au surface owing to the electronegativity of the imidazole ring of the additives and adsorb first on the surface before EC molecules. After the binding reaction of the BZ additives on the Au surface was completed, the EC–Li+ ionic clusters were attracted by the imidazole structure of the BZ additives owing to the Lewis base structure and the repulsion of the dissociation reaction. The repulsion time was approximately 4–10 min. The samples can be ordered as follows in terms of repulsion time: Sample D < Sample B < Sample C. This thus indicates that Sample D had the shortest repulsion time. In addition, with different functional group designs on the BZ additives, the association rate constant demonstrated that the high electro-withdrawing group (-CF3, Sample D) dramatically enhanced the rate of the binding reaction between BZ and the Au surface. Compared with Sample B (-H), the association rate of Sample D increased by 8–9 times. Conversely, the electro-donating group design (-CH3, Sample C) extended the binding reaction time owing to the strong attractive force with Au. As indicated in the Slope II region, Sample A showed a continuing dissociation reaction (reflection angle decreased) of the EC–Li+ cluster corresponding to a decreasing potential. In this region, the electrochemical voltage triggered the decomposition of the EC and formed an SEI on Au. Therefore, the reflection angle change observed for Slope II was lower than that observed for Slope I; this can be attributed to the EC decomposition effect. From the results presented in Table 1, the dissociation rate constant of the EC–Li+ cluster in this reaction range decreased to 32.8 s−1. According to the literature [5,31,32], an SEI consists of several organic and inorganic lithium carbonates from its two-step reactions, indicating that SEI formation drastically influences the dissociation process of EC–Li+ clusters. Notably, three BZ additives displayed individual behaviors in the Slope II region, indicating that the functional group substitutions played critical roles in the electrochemical reaction. With a strong electro-withdrawing group (-CF3) effect in Sample D, the spectrum demonstrated a dissociation reaction process; -CF3 was effectively used to repulse the lithium ion, which dissociated from its original EC–Li+ cluster. Previous research identified, through a simulation, that LiTFB has the lowest dissociation energy and the longest bond length of Li anions compared with other BZ additives [8], implying the lithium ions are easily released from the BZ structure. The dissociated EC was then observed to be decomposed and react with single Li+ to form the SEI after this dissociation process. The dissociation rate constant of Sample D in the Slope II region was calculated to be 48.8 ng−1s−1 (a second-order reaction). Compared with the Slope I region, the reflection angle observed for the Slope II tended to change dramatically (Δθ = 0.063°), indicating that -CF3 effectively inhibited the solvation and association of LiTFB and the EC–Li+ cluster and triggered SEI formation. However, Sample B in the Slope II region demonstrated a similar result to Sample D: the rate constant changed from an association reaction to a dissociation reaction and the reflection angle decreased (a second-order reaction). Nevertheless, without the -CF3 substitution, the electron cloud density of LiBZ was not strong enough to continuously repulse the lithium ions from its original EC–Li+ cluster, owing to the electron hole on the proton (-H) position of the BZ structure. Therefore, the reflection angle change observed for this Slope II dissociation reaction tended to be weaker (Δθ = 0.023°) and lasted a shorter time. This result confirms that SEI formation was
2.3. In operando cells Polyether ether kethone constituted the in operando SPR and QCM cells, which were insulated with an NBR-70/Kalrez type O-ring. A Pt pseudoreference (Pt pseudo) electrode and Pt counter electrode (CE) [24] were used in this study. The Au working electrode, Pt CE, and Pt pseudo electrode were directly connected to a multichannel potentiostat (Biologic VMP3). Before the experiment was initiated, the control sample and modified electrolyte mixtures were injected into the cell (bottom left corner) until the space was filled. Notably, Kalrez O-rings are the best option for making seals because they are resistant to swelling in the presence of organic electrolytes. In this study, we used a Pt electrode as a pseudoreference to compare all the recorded potentials with the Ptpseudo potential. According to a report and an investigation [3,25,26], an EC undergoes a reduction reaction at −1.24 V (equal to 0.76 V vs Li/Li+). For the safety of the EC-SPR compartments, we used diluted 0.1 M LiPF6 in EC:EMC (1:2 in volume) electrolytes for Sample A and 0.1 wt % of BZ salt additive for Samples B, C, and D. 2.4. Electrochemical operations and data analysis In operando QCM measurements were performed using the frequency change of the linear sweep voltammetry (LSV) process (scanning rate, 1 mV s−1; voltage range, 0 to −2 V). Through the Sauerbrey equation, the frequency changes derived from the experimental results were converted into mass changes [27]. We adopted 9 MHz of Au chips, and the theoretical mass sensitivity (ks) was considered to be 18.31 × 107 Hz g−1 cm−2. Δf = -ks × Δm (Δf, frequency changes; ks, theoretical mass sensitivity; and Δm, mass changes). In operando SPR measurements were performed in the angular scanning mode to evaluate angle (θ) changes during the LSV process (scanning rate, 1 mV s−1; voltage range, 0 to −2 V). To analyze the refractive index, thin-film thickness, and angular scanning curves, Winspall software (ver. 3.02) was used to simulate the SPR curves required for calculating the mass and thickness changes. Winspall software is a simulation package that facilitates calculations regarding SPR performance based on Fresnel's equation. This software computes the reflectivity of optical multilayer systems. The thickness and refractive index of the Au layer was determined using a Bionavis™, which was used as a reference for our simulation. Basic kinetic information on the binding rate constant owing to angle changes could also be obtained from sensogram data with the help of TraceDrawer (ver. 1.8.2. X) software. It helped us evaluate, summarize, and report the rate constant and resonance angle changes from the real-time experimental outcomes of physical and chemical interactions at the interface. 3. Results and discussion 3.1. Analysis of in operando SPR Fig. 1 shows the in operando sensogram profiles of all the electrolyte systems in the SPR measurements. After the electrochemical reaction was triggered, an independent dissociation process was observed for sample A (the reflection angle decreased), as indicated by the Slope I region (0–9.5 min) in Fig. 1a. At the stage of electrolyte preparation, LiPF6 underwent dissociation, and the dissociated single lithium ions were solvated by the EC. EC–Li+ ionic clusters were formed due to the high dielectric constant of the EC [3,28–30] with Li+ and diffused close to the Au electrode surface because of the concentration gradient effect. The EC–Li+ ionic clusters were then dissociated to single lithium ions by the electrochemical reaction on the electrode surface. The dissociation rate constant of the EC–Li+ ionic cluster in the Slope I region 428
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Fig. 1. In-operando sensogram profiles of reflection angle of samples (a) A, (b) B, (c) C, and (d) D using 1 mVs-1 as the scanning rate in the electrochemical window of 0 - (−2)V.
15.8 and 16.8 ng−1s−1, respectively. We can conclude that the Li-Au alloying reaction was independent of the effects of LiTFB and LiBZ salt additives. However, Sample C still illustrated a continuous association reaction, indicating that the LiMB salt additive exhibited considerable performance in scavenging the EC–Li+ cluster. In this system, we expected that no more Li-Au alloying layer would be formed. The association rate constant of Sample C in the Slope III region was calculated to be 61.8 s−1. Table 2 shows the simulation results for the forming layer structure on the Au electrode surface after the electrochemical reaction on BZ addition. The thicknesses of the adhesion material, Au, SEI, and Li-Au alloying layers were estimated at 1.53, 49.61, 8.6, and 0.03 nm, respectively, in Sample A. In Samples B, C, and D, the SEI thicknesses were 2.73, 2.60, and 5.20 nm, respectively, which were determined to be consistent with the results shown in Fig. 1d that LiTFB in Sample D had the greatest performance in SEI formation. In addition, the thickness of the Li-Au alloying layer in Sample C was nearly 0 nm, validating the results in Fig. 1c.
triggered in the Slope II region of Sample B, but the reaction level was lower than that of Sample D. The dissociation rate constant of Sample B in the Slope II region was calculated to be 25.6 ng−1s−1 (a second-order reaction). Furthermore, with a strong electron-donating group (-CH3) effect, the Slope II region of Sample C displayed a continuous association reaction, indicating that -CH3 was used to strongly scavenge the lithium ions from the original EC–Li+ cluster. Previous research also identified, through a simulation, LiMB to have the largest dissociation energy and the shortest bond length of Li anions compared with other BZ additives [8], implying that lithium ions are barely released from the BZ structure. We expected that the two integral association reactions in the Sample C system would show the EC is decomposed shortly, revealing that the strong -CH3 substitution effectively inhibited the desolvation and further decomposition of the EC. The association rate constant of Sample C in the Slope II region was calculated to be 70.3 s−1 (a first-order reaction). In the Slope III region, Sample A showed a continuous dissociation reaction, which is consistent with the previous Slope I and II regions, revealing that the lithium ion was dissociated from its original EC–Li+ cluster and joined the Li-Au alloying reaction below −2 V (vs Pt/Li+). In this reaction step, the dissociation rate constant was calculated to be 17.0 s−1, which was much lower than the values in the Slope I and II regions. We supposed that the Li-Au alloying layer formed on the electrode surface when the potential was approached to −2 V. The LiAu alloying reaction disturbed the dissociation of the EC–Li+ cluster and decreased the reaction rate. Samples B and D demonstrated similar results in the Slope III region, in which the rate constant values were
3.2. Analysis of in operando QCM Fig. 2 shows the in operando QCM measurements for Samples A, B, C, and D at voltages ranging from 0 to −2 V and a scanning rate of 1 mVs−1 on the Au working electrode. Fig. 2a presents two reduction reactions: SEI formation at −0.81 V and Li-Au alloying reaction at −1.93 V. Apart from these two reactions, two resting (zero current 429
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Fig. 2. In-operando mass profiles of the samples (a) A, (b) B, (c) C, and (d) D using 1 mVs-1 as the scanning rate in the electrochemical window of 0 - (−2)V.
flows) regions were observed according to the LSV curve. From the mass curve, the reactions could be divided into four regions in terms of the change in the mass curves. The reaction in the first region (0–23 min; 0 to −0.85 V) produced a mass of 1.31 μg cm−2 with zero current behavior, indicating that no electron transfer occurred in this voltage area, which can be correlated with the EC–Li+ cluster adsorption on the Au surface. In the reaction in the second region (23–26 min; −0.85 to −1.13 V), the mass curve showed a sudden increase engendered by the EC decomposition. In this region, the SEI formed a mass of 1.52 μg cm−2 with a nonzero current flow of electrochemical reaction. As the voltage increased, the mass curve continuously increased and tended to approach the second zero current area, which is the third region (26–41 min; −1.13 to −1.62 V). In this region, the reaction was neither an electrochemical reaction nor an EC–Li+ cluster adsorption. This was defined as SEI recombination, because SEI restacking or reconstruction occurred and engendered an increase in density. In this region, SEI recombination generated a mass of 1.31 μg cm−2. When the potential approached −2 V, the reaction in the fourth region (41–50 min; −1.62 to −2 V) involved rapid and highintensity Li-Au alloying formation, indicating the lithium ions dissociated from the EC–Li+ cluster and joined the reaction while the
Table 1 The kinetic parameters measured from in-operando SPR of all samples. Slope
1 2 3
Samples A
B
C
D
Rate constant
Rate constant
Rate constant
Rate constant
73.8 32.8 17.0
89.9 25.6 15.8
85.6 70.3 61.8
774.0 48.8 16.8
Hint: The units of 1st order dissociation/association rate constant (s-1) and 2nd order dissociation rate constant (ng-1 s-1). Table 2 Winspall simulation results for samples A, B, C, and D.
Adhesion layer/nm Au/nm SEI/nm Au-Li alloying/nm
A
B
C
D
1.53 49.61 8.60 0.03
1.53 49.75 2.73 0.03
1.53 50.01 2.60 ∼0
1.53 49.83 5.20 0.04
Table 3 The kinetic parameters measured from in-operando QCM of all samples. Reaction Sample
EC-Li + cluster/μg
EC decomposition/μg
SEI recombination/μg
Li-Au alloying/μg
Total/μg
A B C D
1.31 0.64 0.44 1.42
1.52 0.49 0.43 1.05
1.31 0.80 0.30 2.45
6.56 2.20 2.02 9.92
10.70 3.89 3.43 14.84
430
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Fig. 3. Contact angle measurements of samples (a) A, (b) B, (c) C, and (d) D on Au chip with electrolyte.
potential reached exactly 0 V (Li/Li+). In this region, the Li-Au alloying formation reaction produced a mass of 6.56 μg cm−2. According to these results, Sample A produced a total mass of 10.70 μg cm−2, as shown in Table 3. Regarding the effect of the BZ additives, the in operando QCM measurements showed noteworthy results for the various samples compared with the Sample A system. In Sample D (Fig. 2d), LiTFB was strongly adsorbed on the Au surface owing to its imidazole ring and -CF3 electron-withdrawing group electronegativity properties. After the LiTFB binding reaction on the Au surface was completed, the EC–Li+ ionic clusters were easily attracted by the imidazole structure owing to its Lewis base structure and repulsion of dissociation in a short time range. Therefore, the mass in the first region (0–22 min; 0 to −0.81 V) was calculated to be 1.42 μg cm−2, indicating a greater mass and longer reaction time for EC–Li+ cluster adsorption compared with Sample A. Sample D thus showed the fastest association rate in this reaction region, as shown in Table 1. These results indicate that LiTFB provided a strong interaction with the EC molecule and delayed the dissociation reaction of lithium ions in this zero-current reaction step. As shown in Fig. 2b and c, compared with Sample D, the electron-donating effects in Samples C and B resulted in extremely low mass values for EC–Li+ cluster adsorption on Au, with the values being 0.44 and 0.64 μg cm−2, respectively. This result is consistent with the in operando SPR measurement results that the electron-donating substitutions were not used to enhance lithium ion dissociation because the LiMB salt additive increased the association reaction with the EC–Li+ cluster and Au surface. In the reaction of the second region, with the current effects, the LSV curve obtained for Sample D showed late EC decomposition at −1.14 V with a strong reaction intensity. In addition, EC decomposition resulted in a mass of 1.05 μg cm−2 on Au at 22–29 min and −0.81 to −1.18 V from the mass curve, as shown in Table 3. In this reaction, -CF3 effectively inhibited the solvation and association of imidazole structures and Li+ but triggered EC decomposition. The EC in Sample D showed a strong electrochemical reaction, with the mass curve exhibiting an increasing tendency, as shown Fig. 2d. From the SPR result, the dissociation rate constant of Sample D was the highest. Conversely, Sample C (-CH3) and Sample B (-H) had EC decomposition mass values of 0.43 and 0.49 μg cm−2 on Au, as shown in Fig. 2b and c. As the electron-donating effect of the electrolyte increased, the intensity of EC decomposition decreased, especially in Sample C with a whole and complete association process for the electrochemical reaction, as shown in Fig. 1c. In the reaction in the third region, without the current effects, the mass curve derived for Sample D showed a continuous increase, indicating that the EC decomposition products may had a loose structure and that restacked or recombined to form a new SEI structure on Au. In this reaction, SEI recombination generated a mass of 2.45 μg cm−2 at 29–46 min and −1.18 to −1.83 V. Notably, Samples B and C displayed less mass in SEI recombination for the same reason as the reaction in the second region. The mass values engendered by the SEI recombination reaction were 0.8 and 0.3 μg cm−2 in Samples B and C, respectively. When the potential approached −2 V, the reaction in the fourth region (46–50 min; −1.83 to −2 V) in Sample D involved a more rapid
and higher-intensity Li-Au alloying formation process compared with Sample A, indicating that the -CF3 substitution greatly enhanced the lithium ion dissociation from the EC–Li+ cluster and enabled the ions to join the reaction, when the potential reached exactly 0 V (Li/Li+). In this reaction, the Li-Au alloying formation in Sample D resulted in a mass of 9.92 μg cm−2. According to these results, Sample D produced a total mass of 14.84 μg cm−2 (Table 3). In Samples B and C, the mass values engendered by Li-Au alloying formation were 2.20 and 2.02 μg cm−2, respectively. According to these results, Samples B and C produced a total mass of 3.43 and 3.89 μg cm−2 (Table 3). 3.3. Analysis of contact angle measurement Fig. 3 shows the contact angles of Samples A, B, C, and D with the Au electrode surface. The contact angle measurement was conducted to test electrolyte compatibility with the Au surface and confirm the in operando SPR results. As presented in Fig. 3c, Sample C showed the lowest contact angle (14.0°) with the Au surface, followed by Samples B (31.4°) and D (32.3°). This result implies that Sample C delivered high binding energy with Au, indicating the associated energy was also high. From the in operando SPR results in Fig. 1c, the sensorgram profile showed a continuous association reaction with the electron-donating substitution effect on the additive structure, and the contact angle measurement confirmed this behavior. Compared with Sample D, the electron-withdrawing substitution considerably raised the contact angle that represented the use of LiTFB, which enhanced the repulsion of lithium ions from the EC–Li+ cluster on the Au. 3.4. Surface energy of Au and graphite To confirm the in operando SPR and QCM measurement results, identifying the surface energy of Au and graphite was crucial. A previous study reported that Au has a surface energy of 1610–1790 mN m−1, whereas graphite has an energy of 50.8 mN m−1 [33–35]. As presented in Fig. 3, Sample C demonstrated the lowest contact angle with the Au surface, revealing that the surface energy of the electrolyte was increased by adding LiMB. Conversely, the LiTFB additive with fluorine substitution dramatically reduced the surface energy, which showed high contact angle behavior. Hsieh et al. has revealed that -CF3 group substitution considerably decreases the surface tension due to the F electronegativity property [36]. However, graphite has a lower surface energy than Au, indicating the high surface energy of Sample C renders it adequate for binding with Au but not with graphite. In this study, Sample D containing electron-withdrawing substitution effects effectively decreased the surface energy and was determined to be suitable for binding with graphite. From this conclusion, the contact impedance of Sample D was expected to be lower than that of Sample C in the graphite system. 3.5. Battery performances in MCMB and Si anode half-cells Fig. 4a shows the impedance analysis of MCMB anode half-cells after the first cycle test. According to the results, Sample D showed the 431
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Fig. 4. MCMB half cell electrochemical measurements of (a) the impedance analysis/simulation fitting results after the first cycle and (b) the cycle performance at room temperature.
lowest SEI (Rsei) resistance (1.85 Ω), as determined from an equivalent circuit model simulation, revealing the chemical and physical properties of the formed SEI were notably reinforced by the LiTFB enhancements, such as ionic diffusivity. In addition, the charge transfer resistance (Rct) of the electrochemical reaction demonstrated that the solvation and association between the imidazole structure and EC–Li+ cluster on the MCMB surface could be eliminated and triggered the reaction of the de-solvated EC and lithium ions. However, the electrondonating group substitution considerably affected the impedances of Rsei and Rct on the Sample C system, indicating that in operando SPR and QCM are suitable for predicting battery performance. Fig. 4b shows the cycle performance of all modified battery systems. The results demonstrate that Sample D involving the electron-withdrawing substitution effect provided excellent dissociation of the EC–Li+ cluster, SEI formation, and contact ability with graphite. However, Si plays different rule in dominating the battery performance compares with the graphite. In accordance with the literature [37,38], Si suffers tremendous volume expansion during the electrochemical reaction. The electrolyte contacts new surface and forms SEI continuously. In this study, Sample C may use to inhibit the dissociation of the EC–Li+ cluster in the electrolyte and form less SEI on Si surface. Fig. 5a shows the first charge and discharge curves that the Sample C is able to enhance the coulombic efficiency to 84.8% and achieve the capacity to 2772.9 mAh g−1. In the meantime, the Samples A and D only demonstrate the capacities to 2407.7 and 2722.2 mAh g−1, respectively. With
cycle test, Fig. 5b indicates that the Sample C displays better cycle performance owing to the elimination of SEI on Si surface. 4. Conclusion This is the first study to apply in operando SPR and QCM to investigate electrolyte additive adsorption and solvation processes on Au, EC–Li+ dissociation reactions on Au, the kinetics of EC–Li+ dissociation reactions, SEI formation and Au-Li alloying reactions, and rate constants. Sample A had the thickest SEI layer, followed by Samples D, B, and C. Sample C possessed high surface binding energy as a result of an electron-donating substitution effect within the electrolyte on the Au surface, as well as its high association reaction with the EC–Li+ cluster. Sample D showed considerable Li+ free movement toward and away from the Au surface because of its high reaction constant. Therefore, an electrolyte containing electron-withdrawing effects could potentially improve lithium ion battery performance with graphite anode. Interestingly, Sample C showed the reverse behavior, it is suitable for Si anode in eliminating the SEI formation. In conclusion, the investigations of the stated parameters using in operando SPR and QCM were successfully accomplished under a low electrolyte concentration as well as a tiny amount of additive. These novel measurements processes provide extreme sensitivity, which could not be achieved with conventional electrochemical techniques.
Fig. 5. (a) The 1st cycle and (b) cycle performance of Si anode half cells. 432
Journal of Power Sources 400 (2018) 426–433
T. Alemu et al.
Acknowledgment
[17] J. Lei, L. Li, R. Kostecki, R. Muller, F. McLarnon, J. Electrochem. Soc. 152 (2005) A774–A777. [18] M. Hirayama, N. Sonoyama, M. Ito, M. Minoura, D. Mori, A. Yamada, K. Tamura, J.i. Mizuki, R. Kanno, J. Electrochem. Soc. 154 (2007) A1065–A1072. [19] Y. Miroshnikov, J. Yang, V. Shokhen, M. Alesker, G. Gershinsky, A. Kraytsberg, Y. Ein-Eli, D. Zitoun, ACS Appiled Energy Materials 1 (2018) 1096–1105. [20] H. Jung, K. Gerasopoulos, X. Zhang, R. Ghodssi, J. Phys. Conf. 660 (2015) 012065. [21] A. Senyshyn, M.J. Muhlbauer, K. Nikolowski, T. Pirling, H. Ehrenberg, J. Power Sources 203 (2012) 126–129. [22] D.P. Finegan, M. Scheel, J.B. Robinson, B. Tjaden, I. Hunt, T.J. Mason, J. Millichamp, M.D. Michiel, G.J. Offer, G. Hinds, D.J.L. Brett, P.R. Shearing, Nature Commmunications 6 (2015) 6924. [23] M. Dollé, S. Grugeon, B. Beaudoin, L. Dupont, J.M. Tarascon, J. Power Sources 97–98 (2001) 104–106. [24] R.L. Sacci, J.M. Black, N. Balke, N.J. Dudney, K.L. More, R.R. Unocic, Nano Lett. 15 (2015) 2011–2018. [25] A.M. Haregewoin, E.G. Leggesse, J.C. Jiang, F.M. Wang, B.J. Hwang, S.D. Lin, Electrochim. Acta 136 (2014) 274–285. [26] P. Bach, M. Stratmann, I. Valencia-Jaime, A.H. Romero, F.U. Renner, Electrochim. Acta 164 (2015) 81–89. [27] F.M. Wang, D.T. Shieh, J.H. Cheng, C.R. Yang, Solid State Ionics 180 (2010) 1660–1666. [28] W. Cui, Y. Lansac, H. Lee, S.T. Hong, Y.H. Jang, Phys. Chem. Chem. Phys. 18 (2016) 23607–23612. [29] R. Naejus, R. Coudert, P. Willmann, D. Lemordant, Electrochim. Acta 43 (1998) 275–284. [30] N. Kumar, J.M. Seminario, J. Phys. Chem. C 120 (2016) 16322–16332. [31] F.M. Wang, H.M. Cheng, H.C. Wu, S.Y. Chu, C.S. Cheng, C.R. Yang, Electrochim. Acta 54 (2009) 3344–3351. [32] C.-S. Cheng, F.-M. Wang, J. Rick, Int. J. Electrochem. Sci 7 (2012) 8676–8687. [33] A. Kozbial, Z. Li, C. Conaway, R. McGinley, S. Dhingra, V. Vahdat, F. Zhou, B. D'Urso, H. Liu, L. Li, Langmuir: The ACS Journal of Surfaces and Colloids 30 (2014) 8598–8606. [34] A. Kozbial, F. Zhou, Z. Li, H. Liu, L. Li, Accounts Chem. Res. 49 (2016) 2765–2773. [35] J. Lee, B. Lee, Carbon letters 21 (2017) 107–110. [36] C.T. Hsieh, J.M. Chen, R.R. Kuo, T.S. Lin, C.F. Wu, Appl. Surf. Sci. 240 (2005) 318–326. [37] S.J. Park, H. Zhao, G. Ai, C. Wang, X. Song, N. Yuca, V.S. Battaglia, W. Yang, G. Liu, Journal of the American Society 137 (2015) 2565–2571. [38] K. Schroder, J. Alvardo, T.A. Yersak, J. Li, N. Dudney, L.J. Webb, Y.S. Meng, K.J. Stevenson, Chem. Mater. 27 (2015) 5531–5542.
The author is grateful for the financial support from the Ministry of Science and Technology (MOST) of Taiwan, R.O.C., under grant numbers 102-2221-E-011-016-MY3, 104-3113-E-011-002, 104-2745-8-011001, 105-3113-E-011-002, 105-2628-E-011-005-MY3, 105-2811-E-011017, 106-3113-E-011-001, 106-2923-E-036-002-MY3, and 106-2923-E007-005. References [1] H. Yildirim, A. Kinaci, M.K.Y. Chan, J.P. Greeley, ACS Appl. Mater. Interfaces 7 (2015) 18985–18996. [2] Q. Lv, Y. Liu, T. Ma, W. Zhu, X. Qiu, ACS Appl. Mater. Interfaces 7 (2015) 23501–23506. [3] S.A. Pradanawati, F.-M. Wang, C.-H. Su, J. Power Sources 330 (2016) 127–131. [4] P. Verma, P. Maire, P. Novák, Electrochim. Acta 55 (2010) 6332–6341. [5] S.S. Zhang, J. Power Sources 162 (2006) 1379–1394. [6] M. Xu, L. Zhou, Y. Dong, Y. Chen, J. Demeaux, A.D. MacIntosh, A. Garsuch, B.L. Lucht, Energy Environ. Sci. 9 (2016) 1308–1319. [7] R. Younesi, G.M. Veith, P. Johansson, K. Edström, T. Vegge, Energy Environ. Sci. 8 (2015) 1905–1922. [8] F.M. Wang, S.A. Pradanawati, N.H. Yeh, S.C. Chang, Y.T. Yang, S.H. Huang, P.L. Lin, J.F. Lee, H.S. Sheu, M.L. Lu, C.K. Chang, A. Ramar, C.-H. Su, Chem. Mater. 29 (2017) 5537–5549. [9] X.G. Teng, F.Q. Li, P.H. Ma, Q.D. Ren, S.Y. Li, Thermochim. Acta 436 (2005) 30–34. [10] X. Zhang, P.N. Ross, R. Kostecki, F. Kong, S. Sloop, J.B. Kerr, K. Striebel, E.J. Cairns, F. McLarnon, J. Electrochem. Soc. 148 (2001) A463–A470. [11] L. Niedzicki, P. Oledzki, A. Bitner, M. Bukowska, P. Szczecinski, J. Power Sources 306 (2016) 573–577. [12] S.A. Pradanawati, F.-M. Wang, J. Rick, Electrochim. Acta 135 (2014) 388–395. [13] L. Niedzicki, G.Z. Żukowska, M. Bukowska, P. Szczeciński, S. Grugeon, S. Laruelle, M. Armand, S. Panero, B. Scrosati, M. Marcinek, W. Wieczorek, Electrochim. Acta 55 (2010) 1450–1454. [14] X. Hu, J. Peng, Y. Huang, D. Yin, J. Liu, J. Separ. Sci. 32 (2009) 4126–4132. [15] K. Zhong, X. Xia, B. Zhang, H. Li, Z. Wang, L. Chen, J. Power Sources 195 (2010) 3300–3308. [16] J. Xu, Y. Hu, T. liu, X. Wu, Nanomater. Energy 5 (2014) 67–73.
433