Using Mixed Salt Electrolytes to Stabilize Silicon Anodes for Lithium

Subscriber access provided by UNIV OF SOUTHERN INDIANA

Energy, Environmental, and Catalysis Applications

Using Mixed Salt Electrolytes to Stabilize Silicon Anodes for LithiumIon Batteries via In situ Formation of Li-M-Si Ternaries (M=Mg, Zn, Al, Ca) Binghong Han, Chen Liao, Fulya Dogan, Stephen E. Trask, Saul H. Lapidus, John T. Vaughey, and Baris Key ACS Appl. Mater. Interfaces, Just Accepted Manuscript • DOI: 10.1021/acsami.9b07270 • Publication Date (Web): 18 Jul 2019 Downloaded from pubs.acs.org on July 18, 2019

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

Using Mixed Salt Electrolytes to Stabilize Silicon Anodes for Lithium-Ion Batteries via In situ Formation of Li-M-Si Ternaries (M=Mg, Zn, Al, Ca) Binghong Han,1* Chen Liao,1 Fulya Dogan,1 Stephen E. Trask,1 Saul H. Lapidus,2 John T. Vaughey,1* Baris Key,1* 1. Chemical Sciences and Engineering Division, 2. Advanced Photon Source, Argonne National Laboratory, Lemont, IL 60439, United States Email: [email protected], [email protected], [email protected]

ABSTRACT Replacing traditional graphite anode by Si anode can greatly improve the energy density of lithiumion batteries. However, the large volume expansion and the formation of highly reactive lithium silicides during charging cause the continuous lithium and electrolyte consumption as well as the fast decay of Si anodes. In this work, by adding 0.1 M M(TFSI)x (M = Mg, Zn, Al and Ca) as a second salt into the electrolyte, we stabilize the anode chemistry through in situ formation of Li-M-Si ternary phases during the charging process. First, lithium silicides and magnesium lithium silicides were synthesized as model compounds to investigate the influence of metal doping on the reactivity of lithiated Si. Using solidstate nuclear magnetic resonance spectroscopy, we show that Mg doping can dramatically suppress the chemical reactions between the lithium silicide compounds and common electrolyte solvents. New mixed-salt electrolytes were prepared containing M(TFSI)x as a second salt to LiPF6 and tested in commercially relevant electrodes, which show higher capacity, superior cyclability, and higher coulombic efficiencies in both half-cell and full-cell configurations (except for Zn) when compared with standard electrolytes. Post-electrochemistry characterizations demonstrate that adding M salts leads to the co-insertion of M cations along with Li into Si during the lithiation process, stabilizing silicon anions by forming more stable Li-M-Si ternaries, which fundamentally changes the traditional Li-Si binary chemistry while minimally affecting silicon electrochemical profiles and theoretical capacities. This study opens a new and simple way to stabilize silicon anodes to enable widespread application of Si anodes for lithium-ion batteries.

KEYWORDS Si Anodes, Stability, Lithium-ion Battery, Electrolyte Additive, Mixed Salt Electrolyte, Prelithiation.

I. INTRODUCTION The rapid development of electric vehicles (EVs), plug-in hybrid electric vehicles (HEVs), and portable electronic devices has created a high demand for a next-generation of lithium-ion batteries with higher specific capacities and higher efficiencies. While significant effort has been placed on identifying and creating new cathode materials, advances in anode development have been identified by modeling

1 ACS Paragon Plus Environment

ACS Applied Materials & Interfaces 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 25

as a more direct pathway to increase the energy density of a full electrochemical cell.1 The commercial anode materials currently used, notably graphite, have excellent stability and properties, however their relatively low capacity has become a limiting factor in the next-generation cell development. With this realization, replacing the traditional graphite anode by silicon has been an area of interest for several years, since silicon is abundance on earth, low-cost, while having a theoretical capacity (~3640 mAh/g) approximately 10 times higher than graphite (~370 mAh/g).2-5 In practice, these attributes are offset by issues associated with the large volume expansion that occurs with the reversible formation of various lithium silicides (LS), and the reactivity of the lithiated Si electrode with highly charged Si2-2 and/or Si-4 anions that react to reduce the binders and electrolyte components.6-10 Together these processes combine to reduce the amount of active lithium, reduce free electrolyte solvents, break binder-surface interactions, and contaminate interfacial surfaces with various impurity phases, resulting in very low cycling efficiencies and an unstable solid electrolyte interphase (SEI).2-3, 11-13 One of the main ways identified by researchers to stabilize Si anodes has been to use more compatible and flexible anode binders such as carboxymethylcellulose (CMC), polyacrylic acid (PAA), and lithiated PAA (LiPAA).14-17 In addition, electrolyte additives that form more stable and elastic SEI than the traditional electrolyte-based SEI layers have been investigated in an effort to identify components whose polymeric form withstands the great volume change of Si during operation. This class of materials includes several electrolyte solvents/additives for Si anodes such as the now commonly used fluoroethylene carbonate (FEC), as well as materials adapted from graphite-based lithium-ion systems including vinylene carbonate (VC), lithium bisoxalatoborate (LiBOB), succinic anhydride (SA), and methylene-ethylene carbonate (MEC).18-25 Apart from the use of alternative binders and electrolytes, advances based on modifying and engineering Si materials to evaluate how morphology, nanostructure, and composite material composition interact to relieve material strain and reduce SEI degradation during cycling.3,

26-28

In

addition to these modifications and electrochemical investigations, synthesis and structural studies have recently shown that substituting Mg, Zn, or Al into lithium rich silicides dramatically improves the thermodynamic stability of the fully reduced lithium silicide compounds, which are similar to the products of lithiated Si anodes.29-30 While previous studies have evaluated the fully reduced endmembers, for example lithium insertion into Mg2Si, these types of studies reported rapid capacity fade that was attributed to the disintegration of electrode materials and the Li trapping inside the electrodes on cycling.31 Studies on Si-Al compounds have reported capacities as high as 2400 mAh/ganode (Al0.25Si0.75), but the irreversible formation of a Li-Al-Si ternary crystalline phase prevented reversible Li insertion/extraction cycles and led to a low capacity retention rate, while low concentrations of Al in sputtered Al-Si thin films interestingly had beneficial effects in the overall electrochemistry while suppressing the formation of Li15Si4.32 In general, previous attempts to use Si compounds, intermetallics, or Zintl phases as anodes for lithium-ion batteries have seen limited success due to but not limited to slow Li-ion transport in Li-M-Si phases, extrusion of M, unfavorable phase transformations during


Page 3 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


lithiation/delithiation processes, or poor cycling capacities due to the high content of inactive metals in the anode materials.33 In this study, a novel approach to modify the silicon electrode surface composition is reported based on adding M(TFSI)x (M = Mg, Zn, Al, and Ca) as a second salt into the electrolyte formulation in low concentrations to promote an in situ formation of amorphous Li-M-Si ternary phases during the charging process. Here inert M are chosen so to avoid any anodic or cathodic activity other than insertion into Si to form ternary Zintl phases. These ternary phases, presumably forming first on the surface, protect the underlying lithiated Si phases, reduce side reactions with the electrolyte, and limit the effect of large volume changes by reacting with freshly exposed surfaces. Within this approach, lithium silicides (LS) and Mg-doped lithium silicides (MLS) were synthesized as model compounds to investigate the presence M on the reactivity of lithiated Si in contact with electrolyte solvents. Using solid-state magic angle spinning (MAS) nuclear magnetic resonance (NMR) spectroscopy it was evident that controlled Mg doping dramatically suppresses reactions between the model compounds and common electrolyte solvents (i.e. ethylene carbonate (EC) and ethyl methyl carbonate (EMC)). The stabilization of highly reactive lithiated-Si phases with this new approach is evident from our electrochemical tests on standard commercially relevant electrodes. The electrodes evaluated using mixed salt electrolytes containing Mg(TFSI)2, Zn(TFSI)2, Al(TFSI)3 and Ca(TFSI)2 as a second salt to LiPF6 show higher capacities and superior cyclabilities with significantly improved coulombic efficiencies when compared to standard electrolyte systems (except for Zn, which does not work in full cell configuration). Post electrochemical NMR, X-ray diffraction (XRD), transmission electron microscopy (TEM) and energy dispersive spectrometry (EDS) characterization experiments demonstrated that adding Mg salts promotes the doping of a small concentration of Mg into Si during the lithiation process to form relatively more stable amorphous Li-M-Si ternaries, which fundamentally changes the traditional Li-Si binary chemistry while minimally affecting the electrochemical profiles and capacities. Using small concentrations (0.1 M) of M cations in the electrolyte helps to avoid the complete formation of crystalline Li-M-Si ternaries except at the end of lithiation or M-Si binaries, effectively avoiding any major M extrusion or unfavorable long range binary-ternary phase transformations without an apparent decrease in Li transport properties in the new bulk and interfacial phases.

II. EXPERIMENTAL SECTION II.A. Material Preparation. LS with the aiming Li:Si ratios of 7:3 was synthesized from a stoichiometric mixture of the Li and Si elements. The mixture was first heated to 750 oC in a covered Ta container held for 1 h, then slowly cooled to 700 oC over a 1-hour period and quenched to room temperature. Freshly prepared LS was ground in an agate mortar and pestle for 5 min in an Ar glovebox (with O2 and H2O levels below 0.5 ppm) before mixing with different electrolyte solvents. For reference, no physical changes were observed for the mortar and pestle after mixing.



MLS with the aiming Li:Mg:Si ratios of 2:1:1 was synthesized from a stoichiometric mixture of Li, Mg, and Si elements. The mixture was first heated to 800 oC in a covered Ta container held for 1 h, then slowly cooled to room temperature. Freshly prepared LS was ground in an agate mortar and pestle for 5 min in an Ar glovebox (with O2 and H2O levels below 0.5 ppm) before mixing with different electrolyte solvents. Mg(TFSI)2, Ca(TFSI)2 and Zn(TFSI)2 were bought from Sigma Aldrich and dried in vacuum oven at 160 oC before use. Al(TFSI)3 was synthesized through the reaction between HTFSI and AlCl3 in anhydrous acetonitrile (all ingredients were from Sigma Aldrich). The solvent was evaporated before being dried under vacuum at 80 oC.

II.B. Chemical and Structural Characterizations. MAS-NMR experiments were performed at 7.02 T (300 MHz) on a Bruker Avance III HD spectrometer. NMR spectra were collected at a Larmor frequency of 44.21 MHz at 298K using a 3.2 mm MAS probe. 7Li spectra were acquired at 10 kHz with a rotor synchronized echo pulse sequence (90°--180°--acq), where  1/r. A π/2 pulse width of 2.5 s was used with a sufficiently long pulse recycle delays of 15 s. The spectra were referenced to 1 M LiCl aqueous solution at 0 ppm. 29Si spectra were acquired at 10 kHz with a single pulse measurement, with sufficiently long pulse recycle delays of 1 s. The spectra were referenced to tetramethylsilane at 0 ppm. XRD patterns for the pristine and mixture samples were obtained using Bruker D8 Advance diffractometer equipped with Cu–Kα radiation source (λ = 1.5418 Å). The powders were sealed into the sample holders using the X-ray transparent Kapton tape in the glovebox. The data were processed using Bruker DIFFRAC.SUITE EVA software. High-resolution X-ray diffraction (HRXRD) data were collected by synchrotron powder diffraction using beamline 11-BM-B at the Advance Photon Source (APS), Argonne National Laboratory, with an average wavelength of 0.414 Å for all compounds. Samples were loaded in Kapton capillaries (0.9 mm diameter) and mounted on bases provided by the APS. Two platinum-striped collimating mirrors and a double-crystal Si(111) monochromator were used for the X-ray optics. The data points were collected at room temperature with a step size of 0.001° 2θ and scan speed of 0.01 °/sec. Data were collected while continually scanning the diffractometer 2θ arm. High resolution and short collection time were obtained by using a unique 12-element Si (111) crystal analyzer/detector. Rietveld refinements were performed using TOPAS30. Scanning electron microscopy (SEM) studies were performed at 20 kV on a Hitachi S-4700-II microscope in the Center for Nanoscale Materials at Argonne National Laboratory. Pristine LS and MLS powders were directly sprayed onto the carbon tape for imaging. TEM studies were performed at 200 kV on a JEOL 2100F in the Center for Nanoscale Materials at Argonne National Laboratory. The images were taken using a Gatan Digital Micrograph v2.01 (Gatan Inc.). This TEM is equipped with EDS. The TEM samples were prepared by washing and sonicating the


Page 4 of 25

Page 5 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


post-electrochemical-test Si electrode in dimethyl carbonate (DMC) and then drop cast onto Cu grids in glovebox.

II.C. Electrochemical Testing. In this work, 2032-type coin cells were used to test the electrochemical performance. Several different electrodes were used in this work for half-cell and full-cell tests, which were prepared by Argonne's Cell Analysis, Modeling and Prototyping (CAMP) Facility. The graphite-free high-Si concentration electrodes (Si electrodes) were prepared by laminating Cu foil as the current collector with a slurry containing 80 wt% commercial silicon powders from Paraclete, 10 wt% hard carbon additive (C45), and 10 wt% Lithium Polyacrylate (LiPAA) silicon compatible binder, mixed in NMethyl 2-Pyrrolidone (NMP). The Si electrode has a final loading of 0.96 mg/cm2 and a thickness of 9 µm (not including the Cu foil). The 15%Si+73%Graphite blended electrodes (15%Si electrodes) were prepared in a similar way on the Cu foil, which contains 15 wt% Si, 73 wt% Graphite, and 2 wt% C45, and 10 wt% LiPAA, with a loading of 3.00 mg/cm2 and a thickness of 27 µm (not including the Cu foil).. The NMC532 electrodes were made of 90 wt% LiNi0.5Mn0.3Co0.2O2 from Toda, 5 wt% C45, and 5 wt% Polyvinylidene Fluoride (PVDF) binder, with a loading of 8.98 mg/cm2 and a thickness of 32 µm (not including the Al foil). The Li-rich HE5050 cathodes were made of 92 wt% Li1.2Ni0.15Mn0.55Co0.1O2 from Toda, 4 wt% C45, and 4 wt% PVDF binder, with a loading of 6.06 mg/cm2 and a thickness of 26 µm (not including the Al foil). In half-cell tests, the negative electrodes were Li metal chips and the positive electrodes were the Si electrodes. While in the full-cell tests, the negative electrodes were Si or 15%Si electrodes with the targeting capacity ~ 2 mAh/cm2, and the positive electrodes were NMC532 or Li-rich HE5050 cathodes with the targeting capacity ~ 1.6 mAh/cm2. The separators were Celgard2320. Various electrolyte formulations were used in this study, with their compositions listed in Table 1. The typical half-cell Si-anode testing protocol involves 3 formation cycles, for complete break-in of all crystalline silicon in the electrode, followed by aging cycles for the half cells. The formation cycles consist of constant current discharge step at a rate of C/20 until 0.01 V lower cutoff voltage is reached, immediately followed by a constant potential (voltage-hold) discharge step at 0.01 V until the current drops below C/200 to ensure full lithiation. Here the C rate is the current applied to fully charge or discharge the silicon content to the theoretical capacity in one hour. Then the cell is charged to 1.5 V at a rate of C/20 to complete one cycle. After 3 formation cycles, aging cycles begin with similar discharge (with or without voltage hold) and charge steps between 0.01 and 1.5 V, but at a faster rate of C/3. The typical full-cell testing protocol involves 3 formation cycles followed by Hybrid pulse-power capability cycles (HPPC) and aging cycles with voltage holds at 4.1 V followed by HPPC cycles. The formation cycles consist of constant current charge steps at C/20 (with C to fully charge the Li content of the cell to the theoretical capacity in one hour) until an upper cutoff voltage of 4.1, 4.5, or 4.7 V, immediately followed by a voltage-hold charge step until the current drops below C/50 to ensure full lithiation of the anode. Then the cell is discharged to 3.0 V at a rate of C/20 to complete one cycle. After



3 formation cycles, aging cycles begin with similar charge steps with voltage holds (until C/20 is reached) and discharge steps between 4.1 V and 3.0 V, again with a faster rate of C/3. Hybrid pulse-power capability (HPPC) cycles (not shown in data plots throughout the manuscript) with discharge/charge pulses at the rate of 2C/1.5C for 10 s are used to investigate the impedance change during full-cell electrochemical tests. The HPPC cycles are inserted after 3 formation cycles and after 90 aging cycles. The discharge area specific impedance (ASI) are calculated as ASI = ΔV/ΔI, where ΔV and ΔI are the voltage and current change during the discharge pulse, respectively. For extended-cycling full cell studies up to 270 aging cycles, 2 cells each were resumed to undergo additional aging cycles described above following the 2nd HPPC cycle set. Table 1. Formulations and notations of the electrolytes used in this study. Notations

Components

Gen2

1.2 M LiPF6 in 30 wt% EC + 70 wt% EMC

GenF

Gen2 electrolyte + 10 wt% FEC

GenFM

GenF electrolyte + 0.1 M Mg(TFSI)2

GenFZ

GenF electrolyte + 0.1 M Zn(TFSI)2

GenFA

GenF electrolyte + 0.1 M Al(TFSI)3

GenFC

GenF electrolyte + 0.1 M Ca(TFSI)2

GenFL

GenF electrolyte + 0.1 M LiTFSI

III. RESULTS AND DISCUSSIONS Previously, researchers have found that substituting Mg, Zn, and Al into a lithium rich LS Zintl phase will significantly improve its thermodynamic stability.29-30 However, the effect of metal substitutions on the chemical stability of these reduced compounds in contact with electrolytes has not been reported. In this study, we first used the synthesized crystalline MLS and LS as model compounds of the charged silicon electrodes to investigate whether the substitution of other metal cations into the lithiated Si could reduce its deleterious interactions with the electrolyte solvents. The Scanning electron microscopy (SEM) images and XRD patterns of the pristine lithium silicide (LS) and magnesiated lithium silicide (MLS) model compounds are shown in Figure 1, which have the target stoichiometry of Li7Si3 and Li2MgSi, respectively. The XRD pattern of the pristine LS powders in Figure 1a shows that the LS powder also contains Li12Si7, a slightly more silicon rich phase in the system that results from high temperature lithium evaporation. Similarly, for the pristine MLS powders, the XRD pattern in Figure 1b shows Li2MgSi and a small amount of Li8MgSi6 phases. Although the actual lithiated species formed during the charging of Si anodes could be amorphous,34 previous solid-state NMR studies have shown that the local Li chemical environment in amorphous LS is comparable to those in the crystalline LS.9-10, 35 Therefore, the LS and MLS model compounds synthesized in this work were used as a model system to


Page 6 of 25

Page 7 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


simulate the potential interactions between electrolyte solvents and lithiated Si anodes. It is worth noting that the freshly synthesized LS powders are found to be surprisingly reactive after packed and spun in ZrO2 NMR rotors despite the large particle sized observed by SEM, the section of ZrO2, a geologically stable oxide, in directly contact with the fresh powders blackened after an overnight experiment (see Figure 1a). Staying in ZrO2 rotor for longer time caused the delithiation of pristine LS and finally the formation of a diamagnetic Li phase (see Figure S1 in the Supporting Information). A similar phenomenon was not observed for powders aged in an Ar-glovebox after several weeks. In comparison, the rotor in contact with fresh MLS powders did not undergo any discoloration after an overnight experiment (see Figure 1b). Considering the reactivity of LS and MLS model compounds, the reactivity experiments done in this study were all measured immediately after mixing the pristine LS or MLS powder with different electrolyte solvents analogous to electrochemically generated Zintl phases being constantly exposed to electrolyte solvents in a battery due to volume changes.

Figure 1. XRD results and SEM images of pristine (a) LS and (b) MLS powders, as well as the photos of NMR rotors in contact with the pristine LS and MLS powders after 24 h. The XRD reference peak positions marked by blue clubs and green diamonds in panel (a) represent the Li7Si3 and Li12Si7 phases10, respectively. The XRD reference peak positions marked by red hearts and yellow spades in panel (b) represent the Li2MgSi and Li8MgSi6 phases, respectively.36-37 The rotors in the photos were made of ZrO2 and were initially colorless. The blackish part in the center of the ZrO2 tube is the region directly in contact with LS powders during NMR measurement, while the white parts on two ends are the untouched regions. To investigate the in-situ reactions of model compounds in contact with different common electrolyte solvents, pristine LS and MLS powders were mixed with EC and EMC, respectively, with the volume ratio of 1:1, followed by NMR measurements. Since EC is in solid form at room temperature, the mixture of EC and LS powders were first heated to 325 K inside the NMR probe to let the EC melt and



encompass the LS and MLS powders and then cooled back to room temperature for NMR measurements. Figures 2a and 2b show the 7Li and 29Si MAS NMR results of the mixture of LS with EMC and EC. Compared with the pristine LS, the mixture of LS and EC shows a positive chemical shift of 2.4 ppm in 7Li

NMR, and a significant change in 29Si NMR, which suggests loss of Li+ from the LS compounds

based on literature reports.10 In comparison, the chemical shifts in 7Li and 29Si NMR after mixing LS with EMC are only 0.7 and 1.9 ppm, implying that EMC can be considerably less reactive in contact with LS and can cause less Li loss compared with EC. The biphasic nature of the synthesized MLS likely introduces a lineshape complexity in Figures 2c and 2d as well as the fact that Li-Mg-Si phases resonate at higher frequencies. In addition, we also tested the reactivity of LS with Gen2 electrolyte, which contains both EC and EMC, and the 7Li NMR result shows that mixing LS with Gen2 electrolyte leads to peak shifts similar to the EC case (Figure S2a). After washing the LS+Gen2 mixture with DMC, the 13C and 1H NMR results in Figures S2b and S2c show the presence of insoluble carbonates. Based on the changes in 7Li NMR results in Figure 2a and the deshielding observed in 29Si resonance in Figure 2b, the reactions shown in Scheme 1 is proposed that silicon anions in LS initiate the reaction to lose electrons (and Li, not shown), reducing and decomposing EC and EMC to produce organic lithiated carbonates with or without gas generation (e.g. LiEDC or LiEC/LiBDC, respectively). The reaction of EC with silicon anions was more pronounced spectroscopically, likely due to the prevalence of ring opening reactions. The results from the LS model compound are consistent with previously reported reaction mechanisms.38 On the other hand, the 7Li NMR results in Figure 2c show that mixing MLS with EC only leads to a small peak shifts of ~0.3 ppm compared with the pristine MLS, which are much smaller than the peak shifts after mixing LS with EC in Figure 2a. Similarly, the 29Si NMR shifts after mixing MLS with EC (Figure 2d) are also much smaller than that after mixing LS with EC (Figure 2b). The peak shifts in 7Li and 29Si NMR after mixing MLS with EMC is even smaller. The above results imply that substituting Mg into the LS lattice can effectively stabilize the lithiated silicides against electrolyte solvents. It must be noted that the compositions studied here cannot be generalized to the electrochemically formed domains of highly reduced silicon anions, therefore these results are not meant to be a predictor for complete stability. The goal of our model-compound study is to modulate the instability of the anions with or without Mg doping.


Page 8 of 25

Page 9 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 2. (a) 7Li and (b) 29Si MAS NMR spectra of pristine LS model compound and its mixtures with EMC and EC, respectively, with the volume ratio of 1:1. (c) 7Li and (d) 29Si MAS NMR spectra of pristine MLS model compound and its mixtures with EMC and EC, respectively, with the volume ratio of 1:1. The mixture of LS+EC and MLS+EC have been heated up to 325 K and then cooled back to 298 K.

Scheme 1. Proposed Reactions Based on Characterization Results 2[Si]-2

[Si-Si]-4

2 e- + [Si]2-2

[Si]-2 O 2

O

O

2 e-

O 2 O

O

O O

O

O

O

+ C 2H 4

O

Encouraged by the model compound chemical stability we report for the first time the use of Mg and other metal salts as electrolyte additives to stabilize the cycling of Si electrodes. The central idea is the co-insertion of Mg and Li cations into Si electrode during the lithiation process, forming less-reactive metal-substituted LS species in an in situ fashion and reducing the reactivity of charged Si anodes. 0.1 M Mg(TFSI)2 was added into the baseline GenF electrolytes to formulate the new GenFM electrolyte (see table 1 for the formula and acronyms of all electrolytes used in this study). The half-cell results on Si electrodes fabricated by Argonne’s Cell Analysis, Modeling and Prototyping (CAMP) Facility are shown in Figure 3. For the baseline GenF electrolyte with no Mg salt, the half-cell delithiation capacities start at ~2600 mAh/gSi, and quickly drop to ~2000 mAh/gSi after three formation cycles at C/20. Then



during the 20 fast C/3 cycles the capacities decrease from ~1500 to ~200 mAh/gSi with the coulombic efficiencies dropping from ~96% to ~92%. In comparison, in GenFM electrolyte containing 0.1 M Mg(TFSI)2, the delithiation capacities start from ~2800 mAh/gSi in the first cycle, and are as high as 2650 mAh/gSi even after three formation cycles at C/20. During the 20 fast C/3 cycles, the capacities decrease from ~2200 to ~1650 mAh/gSi and the coulombic efficiencies are maintained at ~97%, which are much better than the baseline GenF electrolyte performance. The capacity drop in GenFM electrolyte from C/20 formation cycles to C/3 aging cycles is likely caused by a rise in cell impedance. If low-cutoff voltage holds at 0.01 V until C/200 is reached at the end of lithiation process are applied not only during the formation cycles but also during the fast aging cycles, a remarkably stable delithiation capacity of ~ 2550 mAh/gSi with coulombic efficiencies ~98.5% can be obtained with the GenFM electrolyte (see Figure S3) for 10-20 cycles until plating starts. In addition, the GenFM half cell shows a faster current fading and stabilization than the GenF cell (see Figure S3c) during potentiostatic holds at 10 mV in the first delithiation process, indicating that GenFM electrolyte reduces parasitic currents by an order of magnitude, which suggests that adding Mg can help to reduce the side reactions of lithiated Si electrodes with electrolytes and therefore improve the cell performance. The approximation can be done due to the fact that the addition of the Mg is the only variable introduced to the system as LiTFSI and/or increased salt concentration has been found to have no effect on performance and Li-M-Si phases has been shown directly to be relatively less reactive from the model compound study. Further ex-situ and in-situ characterization studies are needed to confirm this in detail. The dQ/dV profiles do not show any major changes in the electrochemistry between the GenFM and GenF electrolytes during the initial formation cycles (also see Figure S4 for voltage profiles). When comparing the first formation cycle with the third formation cycle, in GenFM electrolyte, less peak shifts can be observed during both lithiation and delithiation processes compared with those in GenF electrolyte, which is consistent with the much better capacity retention observed in GenFM electrolyte. The dQ/dV profile (i.e. the voltage profile in Figure S4) of the half cell cycled in GenFM electrolyte is totally different from the cycling data on Mg-Si alloy anodes reported previously.31, 33 We observed almost no contribution to the dQ/dV profile from the addition of Mg salt except subtle shifts of the lithiation peak at 0.21 V (to higher voltages) and the delithiation peak at 0.42 V (to lower voltages), which are likely due to small amounts of Mg inclusion into the Zintl phase. These results suggest that adding Mg(TFSI)2 as a second salt (anions other than TFSI- in principle should work just as well) is a simple and effective way to suppress the side reactions between the electrolyte and Si electrodes without diminishing the high capacity of Li-Si chemistry, and at the same time increasing the retention rates and efficiencies of Si anodes, which is consistent with our model-compound observations.


Page 10 of 25

Page 11 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 3. Half-cell electrochemical test results on Si electrodes. The GenF and GenFM electrolytes were used in the electrochemical tests. The delithiation capacities and coulombic efficiencies are shown in panels (a) and (b), respectively. The differential capacities at the first and third formation cycles in GenF and GenFM electrolytes are shown in panels (c) and (d), respectively. The cells were cycled between 0.01 V and 1.5 V, first at the rate of C/20 for 3 formation cycles, then at the rate of C/3 for aging cycles. The capacities are normalized by the mass of silicon. Error bars represents the standard deviations of at least three measurements for each sample. To understand whether Mg is co-inserted into the Si anodes during the lithiation process, fully lithiated electrodes were harvested to obtain the active materials from the Cu foil for further characterization experiments. The harvested electrode powders have been washed with DMC to remove the precipitated Li and Mg salts left on the surface. As shown in Figure 4a, the Si electrode lithiated in GenFM electrolyte (containing 0.1 M Mg(TFSI)2) shows distinct but broad peaks in the XRD pattern, suggesting the formation of Li15-xMgxSi4 phases (assuming a constant 15:4 ratio between cations and Si) with small-crystalline or semi-amorphous feature.29-30 More importantly, the formation of a single Li14.65Mg0.35Si4 phase (Mg:Si ratio of ~1:11) is confirmed by synchrotron High-resolution X-ray diffraction (HRXRD) characterization and refinement analysis shown in Figure 4b. This is a critical observation since it directly proves Mg co-insertion in the bulk and confirming that M incorporation can generate a single phase product electrochemically at full lithiation of silicon. The refinement shows lattice parameters of a = b = c = 10.7211 Å, close to the Li13.63Mg1.37Si4 phase reported before (ICSD 429404, where a = b = c = 10.7418 Å). Synchrotron HRXRD also confirms the minor (042) reflection



at 9.9o (see the insert in Figure 4b) due to ternary formation that was not detected using lab X-rays due to low S/N. No crystalline compound such as Mg2Si was detected by XRD, HRXRD, or TEM (see Figure 4e) after lithiating Si electrode in GenFM electrolyte. However ~3% MgO was estimated from HRXRD as well as other less minor crystalline phases (evident in difference plot), the former likely due to the reaction of Mg(TFSI)2 with water impurity (in the binder and the electrolyte) and/or the silicon oxide shell present on the pristine silicon particles and the latter due to other unconfirmed decomposition products such as MgF2, Mg(OH)2, other electrolyte reduction species etc. The EDS results of the lithiated material in Figure 4e show that significant amounts of Mg can be found both on the surface and in the bulk of lithiated Si particles with the Mg:Si ratio of ~1:9 after the lithiation in GenFM electrolyte, in agreement with the HRXRD results, indicating that using Mg-containing electrolyte does allow for the co-insertion of Mg cations into the Si during the lithiation process, forming semi-crystalline Li-Mg-Si ternaries below the Mg:Si concentration in Li14MgSi4. Meanwhile, the formation Mg-Li or Mg-Si structures are not observed during the in situ Mg insertion/doping process, unlike previously reported cases where higher Mg contents were used in solid state.33 The electrode with 1 full lithiation/delithiation cycle was also studied and its XRD pattern is shown in Figure 4a, which did not indicate any crystalline phases. Figure 4f shows that after delithiation to 1.5 V vs. Li in GenFM electrolyte, the Mg:Si molar ratios increase on the surface and decrease in the bulk compared to the fully lithiated sample. This implies that during the delithiation process, some Mg cations have migrated from the bulk of Si to the surface and/or trapped by the SEI layer. After 9 cycles in GenFM, when the Si electrode is fully lithiated again, Mg can still be found evenly distributed in the surface and bulk of the Si particles, however the Mg:Si ratio drops down to ~1:20 (see Figure S5a). It is interesting to note that if a different silicon batch with more crystalline Si and SiO2 content is used (e.g. SiOx), while an improvement in initial capacity and capacity retention was again noted, significantly more Mg can be found on the surface regions (see Figure S6), consistent with the formation and stability of MgO, Mg being very oxophyllic.39 The formation of MgO could insulate the surface and hinder the Li transportation processes, but it may also help to passivate the reaction between lithiated Si and electrolyte. The clear role of MgO in SEI require further investigations in the future. Si materials from different vendors with different surface oxidation status and/or particles sizes also show various response to the adding of Mg salt, see Figures S6. The smaller Si with fewer surface oxidation such as 80 nm spherical particles from HydroQuebec with only a few nm of SiO2 (see Figure S6d)40 demonstrates considerable capacity and efficiency improvement after adding Mg, while the larger Si with thicker surface oxidation layers (such as NanoAmor or Sigma Aldrich Si in Figure S6e-f) shows even worse performance with GenFM electrolyte. Based on these results such oxophyllic M chemistries by nature are better suited to systems that M can access Si more readily and with binder, solvent, particles and processing methods with less water and oxide content such as SiO2. The Si electrodes lithiated in GenFM electrolyte were also characterized by MAS NMR. The 7Li NMR results in Figure 4c demonstrate the evolution of Li contents and its local environments after


Page 12 of 25

Page 13 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


different number of cycles for fully lithiated electrodes. The peak around 0 ppm is from the diamagnetic Li species mainly in the SEI layer, while the peaks at higher frequencies than 5 ppm represents the Li inserted into the Si. Compared with the electrochemically obtained Li15+xSi4 phase (marked with green dotted line at 6.7 ppm),10 a significant shift of 4.3 ppm in the Li resonance peak of lithiated Si electrode after initial discharge (marked with black dash line at 11.1 ppm) can be observed in Figure 4c. Such a shift of the resonance towards high frequencies is consistent with the doping of Mg cations (with more valence electrons) into the fully lithiated Si electrode. No resonance due to Li14MgSi4 composition was observed (previously reported at 43 ppm29), consistent with the previous EDS and XRD results. After 11 and 17 cycles in GenFM electrolyte, the 7Li NMR results on the re-lithiated Si electrodes show a resonance shift to lower frequencies by -1.7 ppm, appearing at 8.6 ppm (marked with red dashed line), suggesting a decrease in Mg concentration in the Zintl phase. This is consistent with the reduction in Mg:Si ratios after cycling observed in EDS analysis. Furthermore, the 29Si NMR of the lithiated GenFM sample shows a peak center around -55 ppm in Figure 4d. When compared with the 29Si NMR of the electrochemically obtained Li15+xSi4 phase (marked with green dotted line),10 a difference of -127 ppm was observed, which is consistent with the positive shifts observed in 7Li NMR results due to electron shielding/deshielding phenomena for the respective nuclei and again indicates the insertion of Mg into the Li-Si Zintl phase to form a ternary. The peak at -110 ppm (marked with purple dash line) is due to the native SiO2 shell in the silicon starting material. There is an expected and large amount of resonance overlap in Figure 4c and Figure 4d spectra due to the complexity of the SEI and decomposition products as well as active materials and will be the focus of a future study. Overall, the characterization results clearly show the incorporation of small concentrations of Mg in Li-Si chemistry forming ternary LiM-Si phases.



Figure 4. Characterization data of electrochemically treated Si electrodes, lithiated down to 0.01 V vs. Li after different number of cycles, and further delithiated back up to 1.5 V vs. Li, all in half-cell setup using GenFM electrolyte. (a) XRD results of Si electrodes lithiated and then delithiated in GenFM electrolyte. (b) HRXRD and refinement results of Si electrodes lithiated in GenFM electrolyte. The insert shows the zoom-in of the spectra around 10o. (c) 7Li MAS NMR of Si electrodes lithiated in GenFM electrolyte after different number of cycles. (d) 29Si MAS NMR of Si electrodes lithiated in GenFM electrolyte after 0 cycle. Blue dashed line marks diamagnetic Li species in SEI. Black and red dashed lines mark the Li resonance peak centers. Purple dashed line marks the SiO2 species in original Si material. Green dotted line marks the position of electrochemically lithiated Li15+xSi4 phase.10 Blue dashed line marks diamagnetic Li species in SEI. Black and red dashed lines mark the Li resonance peak centers. Green dotted line marks the position of electrochemically lithiated Li15+xSi4 phase.10 (e)(f) TEM and EDS results of Si electrodes lithiated and then delithiated in GenFM electrolyte, respectively. The error bars in EDS results represent the standard deviation of 9 different particles.


Page 14 of 25

Page 15 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Zn(TFSI)2, Al(TFSI)3 and Ca(TFSI)2, were also added into the GenF electrolyte as the secondary salts for Si anodes and tested (the corresponding electrolytes were noted as GenFZ, GenFA, and GenFC, respectively). The half-cell electrochemical test results in Figure S7 show that with Zn- or Al-containing electrolytes, the Si electrodes also show a high delithiation capacity of ~ 2800 mAh/gSi. The Ca-containg GenFC electrolyte shows slightly lower capacity than the GenFM electrolyte, but its capacity performance is still much better than the baseline GenF electrolyte shown in Figure S3. The capacity retention rates and cycling profiles of the half cells with Zn, Al-, or Ca-containing electrolytes are similar if not better than that of the Mg-containing electrolyte. This indicates adding the secondary cations in the electrolyte that can promote the formation of stable ternaries is a simple and effective way in general to stabilize the Si electrodes for Li-ion batteries. We also discover that compared with Mg, the insertion of Ca ions into the Si during the lithiation process is more difficult. As shown in Figure S5b, after the initial lithiation in the GenFC electrolyte, the bulk Ca:Si ratio is only ~1:50, much lower than the Mg:Si ratio of ~1:11 after the initial lithiation in the GenFM electrolyte, which is probably because Ca has a larger cation size. The surface Ca:Si ratio (~1:30) is much higher than the bulk after the initial lithiation, which suggests slower kinetics for Ca insertion and diffusion. The Ca:Si ratio determined by the refinement of the HRXRD result of Si electrode lithiated in the GenFC electrolyte is ~1:33 (see Figure S5d), which is consistentwith the EDS results. After 3 cycles, the surface and bulk Ca:Si ratios slightly increased, as shown by the EDS results in Figure S5c, which are still much lower than the Mg:Si ratios of Si electrodes lithiated in GenFM electrolyte. In addition, adding LiTFSI into the electrolyte (i.e. the GenFL electrolyte) does not lead to similar performance improvements in half-cell tests, as shown in Figure S7, implying the superior performance gained in GenFM, GenFA, GenFZ, and GenFC electrolytes is originated from the adding of secondary cations, rather than the addition of TFSI- anions. However, as will be discussed below, electrochemical compatibility of the secondary cation is critical. Only GenFM, GenFA, and GenFC electrolytes were found to function in full-cell configurations. This is most likely due to the fact that Zn ion is known to be active in the positive electrodes,41 whereas Al ion is known to be just as inactive as Mg and Ca ion.42 In order to evaluate real-world battery performance against two standard baselines (Si+GenF and Graphite+Gen2), the new electrolyte formulations were tested in a full-cell configuration in coin cells using standard commercially relevant electrodes (electrode loading levels given in the experimental section) made in Argonne’s CAMP Facility. Full cells were assembled using NMC532 (LiNi0.5Mn0.3Co0.2O2) or Li-rich HE5050 (Li1.2Ni0.2Co0.2Mn0.6O2) cathodes countered by graphite-free Si anodes or traditional graphite anodes. The full-cell configurations were chosen based on the observations that neither M cation intercalates into graphite nor has an appreciable insertion chemistry related to the selected cathodes. The full-cell electrochemical performance is shown in Figure 5, and the representative voltage profiles are shown in Figure S8. For the combination of NMC532 cathode + Si anode, when aggressively cycled between 3.0 V and 4.1 V at C/3 (with 4.1 V holds at the end of each charge cycle),



using GenFM electrolyte can lead to a higher capacity retention rate of 68% over 90 cycles compared with that using baseline GenF electrolytes (49% over 90 cycles). Such improved capacity retention rate is still lower than the baseline graphite-anode result (96% over 90 cycles), but further developments in optimized and compatible binders and electrolyte formulations as well as new Si materials can certainly improve this performance further. Simply using Si anodes dramatically increases the specific total electrode weight capacity and energy density of the full-cell configuration, as shown in Figures 5b and 5c. Adding a secondary Mg salt into the electrolyte brought an improvement of about 0.15% to the coulombic efficiency (see Figure 5d), which were over 99.3% although still lower than that of graphiteanode full cells (~99.5%). In addition, using Mg-Li mixed salts leads to lower initial impedance after the formation cycles as well as after 90 aging cycles, as shown in Figure S9a, which is likely caused by a more favorable SEI formation on Si electrodes with Mg co-insertion. Doubling the Mg concentration does not provide any obvious benefits to capacity or cyclability (see Figure S10), suggesting enough Mg is provided for the co-insertion process. The fact that no negative response is detected with 0.2 M Mg(TFSI)2, additional salt reservoir can be used to avoid depletion of the secondary salt due to extrusion or other reactions in certain long life applications. Another feature of the new chemistry is that the new formulations are fully compatible with graphitic anodes and can be used with graphite + silicon blends. As shown in Figure S11, the full cells made by 15%Si+73%Graphite blended electrodes vs. NMC532 cathodes gain very similar performance improvements in the GenFM electrolyte, clearly demonstrating that the Mg salt is not impacting the electrochemistry at the graphite anode. On the other hand, if Li-rich layered oxides (TODA HE5050 with 20% additional Li) were used with an increased voltage cutoff to 4.5V during the initial three C/20 formation cycles for the activation and the removal of extra Li,43 the additional Li extracted from the cathode can be used to compensate the greater amount of Li consumption during the initial SEI formation on Si anodes, which is commonly referred as the 1st cycle irreversible capacity loss.44 As a result of such intrinsic prelithiation, much higher capacities can be obtained by using the combination of Li-rich HE5050 cathode + Si anode when again cycled aggressively between 3 and 4.1 V using GenFM electrolyte, as shown in Figure 5e. Here the high-voltage cutoff of the formation cycles should be high enough that enough Li can be extracted to compensate the initial Li consumption, and at the same time not too high to cause the structural damage on cathodes or Li plating on the anode. In this paper, the high-voltage cutoff of the formation cycles was selected as 4.5 V because it can provide better electrochemical performance than the lower (4.1 V) or higher (4.7 V) cutoffs, as shown in Figure S12. Figures 5e-5g show that after adding Mg into the GenF electrolyte, the discharge capacities and energy densities of the HE5050+Si full cells increased by a dramatic ~30%, indicating that adding Mg secondary salt is an effective way to reduce the Li loss during the initial SEI formation with a more abundant Li source. Adding Mg salt also leads to lower initial impedance and slower impedance increase after 90 aging cycles, as shown in Figure S9b. The positive effect of intrinsic prelithiation can also be seen in unusually high initial coulombic efficiency


Page 16 of 25

Page 17 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


values ~99.4% for silicon, approaching levels that can only be obtained with graphite-only anodes. Figure 5g shows the combined effect of using a Si electrode with GenFM electrolyte against a Li-rich electrode on the full cell electrochemical performance. The total active material weight specific energy density surpasses both graphite anode and silicon anode baselines by a remarkable 30% after 90 cycles. The cells tested above show stable extended cycling performance in Figure 6, with ~80 mAh/gcathode capacity after 270 cycles for cells with Li-rich cathodes (33% higher over baseline), and ~55 mAh/gcathode after 270 cycles for cells with NMC532 cathodes (120% higher over baseline). After 270 cycles, the Si electrodes in HE5050+Si full cells cycled in GenFM electrolyte show a bulk Si:Mg ratio of ~24:1, indicating there is still some Mg left inside of the Si particles after long cycling (see Figure S13). Almost no Mg is detected in HE5050 cathode particles after 270 cycles (except for some potentially unwashed Mg salt left on the surface), indicating that Mg has little influence on the cathode. Li rich cathodes can also be activated further (beyond formation cycles) during aging cycles by supplying additional Li to “condition” the Li-diminished cell at the expense of some cathode average voltage, as shown in Figure S14, which could be a further useful feature of this specific anode-cathode couple. The dramatic increase in performance for lithium limited systems such as NMC532 cathodes and the compounded improvement in performance gained via use of mixed salts and intrinsic prelithiation clearly rationalize further improvements in capacities, cyclabilities, and efficiencies that can be obtained by optimizing other cell components such as binders and Si materials.33

Figure 5. Full-cell electrochemical test results. Cells in panels (a)-(d) consist of NMC532 cathodes and Si or Graphite anodes (formation cycles are not shown). Cells in panels (e)-(h) consist of HE5050 cathodes and Si or graphite anodes (formation cycles are not shown). At the end of each charge step the



Page 18 of 25

cells were held at 4.1 V until the current dropped below C/50. The discharge capacities in panels (a) and (c) are normalized by the weights of cathode materials (NMC523 or HE5050). The discharge capacities and discharge energy densities in panels (b), (c), (f), and (g) are normalized by the total weights of cathode materials (NMC523 or HE5050) and anode materials (Si or graphite). Error bars represent the standard deviations of at least three measurements for each sample. Trend lines for the columbic efficiencies in (d) and (h) were calculated by a moving average of 15 adjacent points. GenFA, GenFC, and GenFZ electrolytes with Al, Ca, and Zn salts were also tested in full-cell configurations. In the case of GenFZ electrolyte, the full cell loses all capacity promptly during the first formation cycle of the cycling protocol (data not shown), which is most likely due to the activity of Znion in the cathode.41 On the other hand, the Al- and Ca-salts like Mg work compatibly in the full-cell configuration. As shown in Figure 5, using Al-containing GenFA electrolyte leads to a lower coulombic efficiency compared with other electrolytes. GenFA also causes fast impedance rising over cycles, as shown in Figure S9. When paired with NMC532 cathodes, GenFA leads to lower initial capacity but better capacity retention than the GenF baseline, which let it outperform the GenF electrolyte after ~80 cycles. When paired with HE5050 cathodes, GenFA shows the initial capacity even higher than GenFM but with a worse capacity retention, which might be due to its high capacity utilization or fast rising impedance. Extended cycles for GenFA show more decay in cyclability and performance when compared to GenFM (see Figure 6). It must be noted that despite the best efforts, GenFA electrolyte in pristine state had a yellow tint that could be due to impurities originating from the salt or indicative of some degradation, which may lead to the low stability and high impedance in GenFA electrolyte. Meanwhile, adding Ca salt into the electrolyte (i.e. using GenFC electrolyte) brings similar improved capacities, retention rates, coulombic efficiencies and impedance as in the GenFM electrolyte in NMC532-based full-cell tests, as shown in Figures 5a-5d and S9a. In HE5050-based full-cell tests, GenFC shows much higher capacities, better cyclability, higher efficiencies and lower impedance than GenFM (see Figures 5e-5h and S9b), indicating that the Ca-containing electrolyte performs better than the Mg-containing electrolyte with initial high-voltage activation process. These GenFC cells show ~70% capacity retention after 270 cycles in extended full-cell tests as shown in Figure 6b. The differences in electrochemical performance trends observed between GenFM, GenFA and GenFA can be due to a number of factors originating from the complexity of the battery systems. In fact, both the cation and anion of the secondary salt can have effects in addition to the starting Si material, the binder and the solvent choices, which all influence the final electrochemical performance of silicon full cell. Therefore, more fundamental studies are critically required in the future for an in-depth understanding of the new chemistries demonstrated in this paper, which perhaps for the first time allow for an effective optimization of the full cell compositions with silicon anodes that can electrochemically outperform the well optimized graphite electrodes over the last few decades.


Page 19 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 6. Discharge capacities of full cells using GenF, GenFM, GenFA, and GenFC electrolytes with extended cycles. (a) The full cells consist of NMC532+Si electrodes were cycled between 3.0 and 4.1 V at C/3 after three formations cycles between 3.0 and 4.1 V at C/20 (formation cycles are not shown). (b) The full cells consist of HE5050+Si electrodes were cycled between 3.0 and 4.1 V at C/3 after three formations cycles between 3.0 and 4.5 V at C/20 (formation cycles are not shown). At the end of each charging process during aging cycles the cells were held at 4.1 V until the current dropped below C/50. Two cells were tested for extended cycles on each cathode to show that the lifetime of the full cells are repeatable.

IV. CONCLUSIONS In this study, we show Li-M-Si ternaries to be chemically stable against common electrolyte solvents and they can be formed in-situ through electrochemical co-insertion after adding M(TFSI)x (M = Mg, Zn, Al, and Ca) as the secondary salts into the electrolyte formulation in low concentrations. Li-M mixed salts stabilize the lithiated Si phases and reduce their side reactions with the electrolytes. The half-cell electrochemical test results show higher capacities, superior cyclabilities, and improved coulombic efficiencies with the new silicon electrolyte formulations containing Mg(TFSI)2, Zn(TFSI)2, Al(TFSI)3 or Mg(TFSI)2 as the secondary salts to LiPF6 in comparison to the standard electrolyte. The postelectrochemistry NMR, (HR)XRD, TEM and EDS characterizations demonstrate that adding Mg secondary salt promotes the doping of small concentrations up to 0.09 Mg per Si into silicon in the bulk during the lithiation process to form relatively more stable amorphous or semi-crystalline Li-M-Si ternaries, which fundamentally changes the traditional Li-Si binary chemistry while minimally effecting the electrochemical profiles, capacities and rate capabilities. Using Mg-containing electrolyte in NMC full cells improves the coulombic efficiencies by 0.15% and in turn the capacity retention rates by ~40%



after 90 cycles. The impedance characteristics are also improved after adding the Mg second salt. We also show that Li-rich cathodes can be used effectively to couple with silicon electrodes for intrinsic prelithiation via the activation of the cathode at high voltages during formation cycles or even after extended cycling, which can dramatically increase the usable silicon capacities on top of the gains in coulombic efficiencies and retention rates, surpassing full cell energy densities of cells coupled to graphite electrodes after 90 cycles. Extended-cycle tests show performance retention and Mg activity even after 270 cycles. Similar full-cell performance improvements can also be found via the addition of an Al or Ca salt into the electrolyte, however the Zn salt is found to be not suitable in full-cell configuration mainly due to the activity of Zn in the cathode. The results also suggest that a rich Zintl chemistry must be explored fundamentally further with this approach where other M cations, particularly the multivalents that can form Li-M-Si Zintl phases with minimum activity in the cell otherwise, can be incorporated via the electrolyte to provide in situ stabilization via the formation of Li-M-Si ternaries, Li-M-M’-Si quaternaries, or higher. The battery chemistry demonstrated in this study introduces a whole new and synergistic approach to stabilize silicon anodes and has the potential to be a fundamental building block in widespread application of Si anodes in lithium-ion batteries, particularly when used in conjunction with the advances in binder, electrolyte, and Si material developments as well as new prelithiation methods.

ASSOCIATED CONTENT Supporting Information The Supporting Information is available free of charge on the Publications website at: Additional TEM, NMR, and electrochemical test results.

Author Information Corresponding Author *Email: [email protected], [email protected], [email protected] Notes The authors declare no competing financial interest.

ACKNOWLEDGEMENT The authors would like to thank Brian Cunningham and David Howell from the Vehicle Technologies Program, at the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy for their support. The work at Argonne National Laboratory was supported by the U.S. Department of Energy, Office of Vehicle Technologies, under Contract No. DEAC02-06CH11357. The submitted manuscript has been created by UChicago Argonne, LLC, Operator of Argonne National Laboratory (“Argonne”). Argonne, a U.S. Department of Energy 20 ACS Paragon Plus Environment

Page 20 of 25

Page 21 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Office of Science laboratory, is operated under Contract No. DE-AC02-06CH11357. The work at Argonne National Laboratory was supported by the U.S. Department of Energy, Office of Vehicle Technologies, under Contract No. DE-AC02-06CH11357. The electrodes in this article were fabricated at Argonne's Cell Analysis, Modeling and Prototyping (CAMP) Facility. This work used SEM and TEM in the Center for Nanoscale Materials, and HRXRD in Sector 11BM in the Advanced Photon Source, all at Argonne, which are supported by the U. S. Department of Energy, Office of Science, Office of Basic Energy Sciences, also under Contract No. DE-AC02-06CH11357. The authors thank Dr. Bob Jin Kwon and Dr. Yanjie Cui at Argonne for experimental support.



REFERENCES (1) Liu, J.; Bao, Z.; Cui, Y.; Dufek, E. J.; Goodenough, J. B.; Khalifah, P.; Li, Q.; Liaw, B. Y.; Liu, P.; Manthiram, A.; Meng, Y. S.; Subramanian, V. R.; Toney, M. F.; Viswanathan, V. V.; Whittingham, M. S.; Xiao, J.; Xu, W.; Yang, J.; Yang, X.-Q.; Zhang, J.-G. Pathways for Practical High-Energy Long-Cycling Lithium Metal Batteries. Nature Energy 2019, 4, 180– 186. (2) Kasavajjula, U.; Wang, C.; Appleby, A. J. Nano- and Bulk-Silicon-Based Insertion Anodes for Lithium-Ion Secondary Cells. Journal of Power Sources 2007, 163, 1003-1039. (3) Wu, H.; Cui, Y. Designing Nanostructured Si Anodes for High Energy Lithium Ion Batteries. Nano Today 2012, 7, 414-429. (4) Maranchi, J. P.; Hepp, A. F.; Kumta, P. N. High Capacity, Reversible Silicon Thin-Film Anodes for Lithium-Ion Batteries. Electrochemical and Solid-State Letters 2003, 6, A198-A201. (5) Winter, M.; Besenhard, J. O.; Spahr, M. E.; Novák, P. Insertion Electrode Materials for Rechargeable Lithium Batteries. Advanced Materials 1998, 10, 725-763. (6) Nesper, R. Structure and Chemical Bonding in Zintl-Phases Containing Lithium. Prog. Solid State Chem. 1990, 20, 1-45. (7) Nesper, R.; Von Schnering, H. G. Lithium Silicide (Li21si5), a Zintl Phase as Well as a Hume-Rothery Phase. J. Solid State Chem. 1987, 70, 48-57. (8) Nesper, R.; Von Schnering, H. G.; Curda, J. Lithium Silicide, Li12si7, a Compound with Trigonal Planar Si4 Clusters and Isometric Si5 Rings. Chem. Ber. 1986, 119, 3576-90. (9) Key, B.; Morcrette, M.; Tarascon, J. M.; Grey, C. P. Pair Distribution Function Analysis and Solid State Nmr Studies of Silicon Electrodes for Lithium Ion Batteries: Understanding the (De)Lithiation Mechanisms. Journal of the American Chemical Society 2011, 133, 503-512. (10) Key, B.; Bhattacharyya, R.; Morcrette, M.; Seznec, V.; Tarascon, J. M.; Grey, C. P. RealTime Nmr Investigations of Structural Changes in Silicon Electrodes for Lithium-Ion Batteries. Journal of the American Chemical Society 2009, 131, 9239-9249. (11) Yoshio, M.; Wang, H.; Fukuda, K.; Umeno, T.; Dimov, N.; Ogumi, Z. Carbon-Coated Si as a Lithium-Ion Battery Anode Material. Journal of The Electrochemical Society 2002, 149, A1598-A1603. (12) Hatchard, T. D.; Dahn, J. R. In Situ Xrd and Electrochemical Study of the Reaction of Lithium with Amorphous Silicon. Journal of The Electrochemical Society 2004, 151, A838A842. (13) Winter, M.; Besenhard, J. O. Electrochemical Lithiation of Tin and Tin-Based Intermetallics and Composites. Electrochimica Acta 1999, 45, 31-50. (14) Hochgatterer, N. S.; Schweiger, M. R.; Koller, S.; Raimann, P. R.; Wöhrle, T.; Wurm, C.; Winter, M. Silicon/Graphite Composite Electrodes for High-Capacity Anodes: Influence of Binder Chemistry on Cycling Stability. Electrochemical and Solid-State Letters 2008, 11, A76A80. (15) Kim, J. S.; Choi, W.; Cho, K. Y.; Byun, D.; Lim, J.; Lee, J. K. Effect of Polyimide Binder on Electrochemical Characteristics of Surface-Modified Silicon Anode for Lithium Ion Batteries. Journal of Power Sources 2013, 244, 521-526. (16) Koo, B.; Kim, H.; Cho, Y.; Lee, K. T.; Choi, N.-S.; Cho, J. A Highly Cross-Linked Polymeric Binder for High-Performance Silicon Negative Electrodes in Lithium Ion Batteries. Angewandte Chemie International Edition 2012, 51, 8762-8767. (17) Magasinski, A.; Zdyrko, B.; Kovalenko, I.; Hertzberg, B.; Burtovyy, R.; Huebner, C. F.; Fuller, T. F.; Luzinov, I.; Yushin, G. Toward Efficient Binders for Li-Ion Battery Si-Based Anodes: Polyacrylic Acid. ACS Applied Materials & Interfaces 2010, 2, 3004-3010. (18) Dalavi, S.; Guduru, P.; Lucht, B. L. Performance Enhancing Electrolyte Additives for Lithium Ion Batteries with Silicon Anodes. Journal of The Electrochemical Society 2012, 159, A642-A646. 22 ACS Paragon Plus Environment

Page 22 of 25

Page 23 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(19) Markevich, E.; Salitra, G.; Aurbach, D. Fluoroethylene Carbonate as an Important Component for the Formation of an Effective Solid Electrolyte Interphase on Anodes and Cathodes for Advanced Li-Ion Batteries. ACS Energy Letters 2017, 2, 1337-1345. (20) Etacheri, V.; Haik, O.; Goffer, Y.; Roberts, G. A.; Stefan, I. C.; Fasching, R.; Aurbach, D. Effect of Fluoroethylene Carbonate (Fec) on the Performance and Surface Chemistry of SiNanowire Li-Ion Battery Anodes. Langmuir 2012, 28, 965-976. (21) Choi, N.-S.; Yew, K. H.; Kim, H.; Kim, S.-S.; Choi, W.-U. Surface Layer Formed on Silicon Thin-Film Electrode in Lithium Bis(Oxalato) Borate-Based Electrolyte. Journal of Power Sources 2007, 172, 404-409. (22) Han, G.-B.; Ryou, M.-H.; Cho, K. Y.; Lee, Y. M.; Park, J.-K. Effect of Succinic Anhydride as an Electrolyte Additive on Electrochemical Characteristics of Silicon Thin-Film Electrode. Journal of Power Sources 2010, 195, 3709-3714. (23) Chen, L.; Wang, K.; Xie, X.; Xie, J. Effect of Vinylene Carbonate (Vc) as Electrolyte Additive on Electrochemical Performance of Si Film Anode for Lithium Ion Batteries. Journal of Power Sources 2007, 174, 538-543. (24) Ma, L.; Glazier, S. L.; Petibon, R.; Xia, J.; Peters, J. M.; Liu, Q.; Allen, J.; Doig, R. N. C.; Dahn, J. R. A Guide to Ethylene Carbonate-Free Electrolyte Making for Li-Ion Cells. Journal of The Electrochemical Society 2017, 164, A5008-A5018. (25) Bridel, J.-S.; Grugeon, S.; Laruelle, S.; Hassoun, J.; Reale, P.; Scrosati, B.; Tarascon, J.M. Decomposition of Ethylene Carbonate on Electrodeposited Metal Thin Film Anode. Journal of Power Sources 2010, 195, 2036-2043. (26) Ma, D.; Cao, Z.; Hu, A. Si-Based Anode Materials for Li-Ion Batteries: A Mini Review. Nano-Micro Letters 2014, 6, 347-358. (27) Su, X.; Wu, Q.; Li, J.; Xiao, X.; Lott, A.; Lu, W.; Sheldon Brian, W.; Wu, J. Silicon-Based Nanomaterials for Lithium-Ion Batteries: A Review. Advanced Energy Materials 2013, 4, 1300882. (28) Nitta, N.; Wu, F.; Lee, J. T.; Yushin, G. Li-Ion Battery Materials: Present and Future. Materials Today 2015, 18, 252-264. (29) Baran, V.; van Wüllen, L.; Fässler, T. F. Substitution of Lithium for Magnesium, Zinc, and Aluminum in Li15si4: Crystal Structures, Thermodynamic Properties, as Well as 6li and 7li Nmr Spectroscopy of Li15si4 and Li15−Xmxsi4 (M=Mg, Zn, and Al). Chemistry – A European Journal 2016, 22, 6598-6609. (30) Zeilinger, M.; Baran, V.; van Wüllen, L.; Häussermann, U.; Fässler, T. F. Stabilizing the Phase Li15si4 through Lithium–Aluminum Substitution in Li15–Xalxsi4 (0.4 < X < 0.8)— Single Crystal X-Ray Structure Determination of Li15si4 and Li14.37al0.63si4. Chemistry of Materials 2013, 25, 4113-4121. (31) Kim, H.; Choi, J.; Sohn, H. J.; Kang, T. The Insertion Mechanism of Lithium into Mg2si Anode Material for Li‐Ion Batteries. 1999, 146, 4401-4405. (32) Fleischauer, M. D.; Obrovac, M. N.; Dahn, J. R. Al–Si Thin-Film Negative Electrodes for Li-Ion Batteries. 2008, 155, A851-A854. (33) Obrovac, M. N.; Chevrier, V. L. Alloy Negative Electrodes for Li-Ion Batteries. Chemical Reviews 2014, 114, 11444-11502. (34) Antartis, D.; Dillon, S.; Chasiotis, I. Effect of Porosity on Electrochemical and Mechanical Properties of Composite Li-Ion Anodes. Journal of Composite Materials 2015, 49, 1849-1862. (35) Trill, J.-H.; Tao, C.; Winter, M.; Passerini, S.; Eckert, H. Nmr Investigations on the Lithiation and Delithiation of Nanosilicon-Based Anodes for Li-Ion Batteries. Journal of Solid State Electrochemistry 2011, 15, 349-356. (36) Nesper, R.; Curda, J.; Von Schnering, H. G. Li8mgsi6, a Novel Zintl Compound Containing Quasi-Aromatic Si5 Rings. Journal of Solid State Chemistry 1986, 62, 199-206. (37) Herbst, J. F.; Meyer, M. S. Structural, Electronic, and Hydriding Properties of Li2mgsi. Journal of Alloys and Compounds 2010, 492, 65-68. 23 ACS Paragon Plus Environment


(38) Michan, A. L.; Leskes, M.; Grey, C. P. Voltage Dependent Solid Electrolyte Interphase Formation in Silicon Electrodes: Monitoring the Formation of Organic Decomposition Products. Chemistry of Materials 2016, 28, 385-398. (39) Wang, H.; Senguttuvan, P.; Proffit, D. L.; Pan, B.; Liao, C.; Burrell, A. K.; Vaughey, J. T.; Key, B. Formation of Mgo During Chemical Magnesiation of Mg-Ion Battery Materials. 2015, 4, A90-A93. (40) Zhang, L.; Liu, Y.; Key, B.; Trask, S. E.; Yang, Z.; Lu, W. Silicon Nanoparticles: Stability in Aqueous Slurries and the Optimization of the Oxide Layer Thickness for Optimal Electrochemical Performance. ACS Applied Materials & Interfaces 2017, 9, 32727-32736. (41) Pan, C.; Nuzzo, R. G.; Gewirth, A. A. Znalxco2–Xo4 Spinels as Cathode Materials for Non-Aqueous Zn Batteries with an Open Circuit Voltage of ≤2 V. Chemistry of Materials 2017, 29, 9351-9359. (42) Lin, M.-C.; Gong, M.; Lu, B.; Wu, Y.; Wang, D.-Y.; Guan, M.; Angell, M.; Chen, C.; Yang, J.; Hwang, B.-J.; Dai, H. An Ultrafast Rechargeable Aluminium-Ion Battery. Nature 2015, 520, 324–328. (43) Dogan, F.; Croy, J. R.; Balasubramanian, M.; Slater, M. D.; Iddir, H.; Johnson, C. S.; Vaughey, J. T.; Key, B. Solid State Nmr Studies of Li2mno3 and Li-Rich Cathode Materials: Proton Insertion, Local Structure, and Voltage Fade. Journal of The Electrochemical Society 2015, 162, A235-A243. (44) Dose, W. M.; Maroni, V. A.; Piernas-Muñoz, M. J.; Trask, S. E.; Bloom, I.; Johnson, C. S. Assessment of Li-Inventory in Cycled Si-Graphite Anodes Using Lifepo4 as a Diagnostic Cathode. 2018, 165, A2389-A2396.


Page 24 of 25

Page 25 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


TOC GRAPHIC


Using Mixed Salt Electrolytes to Stabilize Silicon Anodes for Lithium

Recommend Documents