Glyme-based liquid – solid electrolytes for lithium metal batteries † 2019, 7 , 13331

The development of stable electrolytes for lithium metal batteries is urgently required. In this work, nanoporous alumina/lithium salt-containing glyme liquid – solid composite electrolytes are proposed. The bene ﬁ cial interfacial e ﬀ ect and transport properties such as ionic conductivity and lithium transference number can be tuned by variation of the cell temperature and electrolyte salt concentration. In symmetrical lithium/electrolyte/lithium cells, glyme-based liquid – solid electrolytes show superior performance compared to their liquid counterparts. However, even in the high lithium transference number cases, unfavorable mossy lithium deposits can be observed. to Fig. 6 Micrographs of the lithium deposit formed during the stripping – platting experiment in Li|AAO: (1 M LiTf/triglyme)|Li cell experiments at 0.3 mA cm (cid:1) 2 current density: (a and b) scanning electron micrographs of the lithium deposit on the surface of porous AAO (c) corresponding energy-dispersive X-ray spectroscopy (EDS) elemental mapping images. The element under study corresponds to the lighter part of the EDS recorded, (d and e) transmission electron micrographs of the lithium deposit. Clear lattice fringes at higher magni ﬁ cation con ﬁ rm


Introduction
Ever since the discovery of the lithium ion batteries, lithium metal has been envisaged as an optimal anode due to the highest theoretical capacity (3860 mA h g À1 ) and the lowest electrochemical potential (À3.04 V vs. hydrogen electrode). [1][2][3][4][5] On the other hand, lithium metal is particularly interesting for the development of the so-called next generation Li-S and Li-O 2 battery technologies. 6,7 Key challenges in lithium metal batteries are poor safety and cycleability, a consequence of the instable solid electrolyte interphase (SEI) forming between the typical electrolytes and the lithium anode, as well as deposition of lithium in mossy or dendritic form. 8,9 Recently, a number of solutions have been proposed, including chemical or physical pre-treatment of the lithium metal to form an articial SEI, [10][11][12][13][14] homogenization of the Li-ion ux, [15][16][17] tuning of the shear modulus of the electrolyte and application of external pressure, [18][19][20][21][22] as well as electrolyte composition adaptation. [23][24][25][26][27] An interesting class of materials, suggested to be useful in lithium metal batteries, are liquid/solid (hybrid) electrolytes, created by inltration of salt-in-solvent electrolytes into nanoporous solid matrix. [28][29][30][31] Appropriate liquid/solid electrolytes offer high ionic conductivities accompanied with good adhesion properties, and high shear modulus. Solids which offer benecial surface adsorption properties are able to immobilize the anions on the surface, locally enhance the Li + conductivity and boost the lithium transference number without compromising overall conductivity. [32][33][34] Such liquid/solid electrolytes have the potential to enable dendrite-free lithium metal batteries at lower currents. 35,36 In this work, we report on anodic aluminium oxide (AAO) membranes inltrated with glyme-based lithium electrolytes as potential candidates for lithium metal batteries. AAO membranes are chosen as they offer well-dened porosities with small enough pore size, while glyme-based electrolytes already showed good stability vs. lithium electrode. 34,37 We investigate systematically how the changes in temperature and salt concentration inuence interfacial effects and transport properties such as ionic conductivity and lithium transference number. Finally, we investigate the failure mechanism of the Li|liquid/solid electrolyte|Li cells as well as the structure of the lithium deposits formed during lithium/stripping plating experiments.

Experimental
Characterization of the anodic aluminum oxide (AAO) The morphology of the AAO monoliths (diameter, d ¼ 13 mm, thickness t ¼ 60 mm, Whatman Anodisc® purchased from Sigma-Aldrich) was observed using scanning electron microscopy (SEM, Zeiss Merlin device operating at 5 kV). The samples were coated with a conducting carbon layer prior to the SEM measurement. The X-ray diffraction pattern was obtained in ambient environment using a Philips device with Cu Ka 1 (l ¼ 0.154 nm) radiation. Magic angle spinning (MAS) NMR measurements were performed on a Bruker Avance III 400 MHz instrument in a magnetic eld of 9.4 T, with a Larmor frequency of 400.1 Hz for 1 H at room temperature and a spinning speed of 12 500 Hz. A N 2 adsorption-desorption apparatus (BELSORP-mini II, Bel Japan Inc.) was employed to characterize the porosity of the AAO membranes. Prior to the measurement, samples were degassed overnight at 120 C under vacuum. Brunauer-Emmet-Teller method was used to calculate the surface area from the N 2 desorption curve. The ion exchange capacity (IEC) of the -OH groups in AAO was The cross-section of the pores can be seen in the center and on the right side, (b) X-ray diffraction pattern of one membrane revealing the amorphous structure of the material, (c) 1 H MAS-NMR spectra of a crushed membrane (black line) was fitted (dotted lines) to estimate the total number of available surface -OH groups. The inset shows voltage evolution for short polarization times, (b) impedance spectra recorded before (black) and after (red) the galvanostatic polarization. Observed semicircle corresponds to the unchanged solid electrolyte interphase (SEI) resistance, (c) extraction of the salt diffusion coefficient from the long polarization times. calculated from the potentiometric titrations. For that purpose, a dened amount of crushed AAO particles was dispersed in a 0.1 M NaCl aqueous solution, stirred continuously for 1 h, and subsequently titrated by a 0.01 M NaOH solution, followed by back-titration with a 0.01 HCl solution up to the neutral pH values.

Electrolyte preparation
The composite electrolytes were prepared by inltration of dry nanoporous AAO with a solution of 0.02-1 M lithium tri-uoromethanesulfonate (LiCF 3 SO 3 , Sigma-Aldrich, 99.995%, LiTf) in tri(ethylene glycol)dimethyl ether (M w ¼ 178.2 g mol À1 , Sigma-Aldrich, triglyme) overnight. Triglyme solvent was distilled prior to use, and the water content of the nal electrolyte thus kept below 150 ppm.

Electrolyte characterization
For the electrochemical measurements, the electrolytes were sandwiched between two symmetric lithium electrodes in a self-designed copper-plated polytetrauorethylen (PTFE) cell. Impedance spectroscopy was applied before and aer galvanostatic polarization in the frequency range from 10 À1 to 10 7 Hz (0.1 V amplitude) using a Solartron 1260 frequency analyzer. When needed, the temperature was externally controlled by a RC6CP Lauda thermostat. The ionic conductivities were calculated from the cell geometry and the obtained resistances. The lithium plating/stripping experiments and the galvanostatic polarization measurements were performed with a Keithley 2604B source-meter instrument. Thermal properties of the electrolytes were determined using a differential scanning calorimeter (DSC 214, Polyma, Netzsch) in the temperature range from À100 C to 80 C at a heating rate of 10 C min À1 under continuous nitrogen purging rate of 60 mL min À1 . Zeta potential was measured at room temperature in a closed vessel, by measuring the magnitude and phase of the Colloid Vibration Current (CVI) at 3 MHz, using an electroacoustic spectrometer (DT-1200, Dispersion Technology, Inc., Quantachrome). Electrolyte preparation and the electrochemical measurements (excluding the temperature dependent measurements and zeta potential measurements) were performed in a glovebox (<0.1 ppm H 2 O, #0.1 ppm O 2 ) under Ar atmosphere.

Lithium dendrite characterization
Upon disassembling the cell in the Ar-lled glovebox, visible brown discolorations (dendrites) inside and on the top of AAO were collected and either deposited on scanning electron microscopy (SEM) sample holders or crushed, dispersed in ethanol and subsequently deposited on carbon grids for transmission electron microscopy (TEM) measurements. TEM measurements were performed using a Philips, CM200 instrument at low energies.

Results and discussion
In the present study, AAO membranes with high pore density, controlled pore size, shape and orientation as illustrated in the SEM image in Fig. 1a, have been employed as the solid part of the liquid-solid electrolyte. The pores of the AAO membrane have conical frustum form with an average upper base radius of 20 nm and a 200 nm lower base radius, Fig. 1a, le. Using the given geometry and the approximate number of pores observed in the SEM micrographs, a surface area value of A calc ¼ 35 m 2 g À1 can be estimated (available in the ESI †). Additionally, the Brunauer-Emmet-Teller (BET) surface area is measured to be of a similar order of magnitude, S BET ¼ 24 m 2 g À1 . X-ray diffraction patterns of the AAO membrane asserts the amorphous nature of the Al 2 O 3 material, Fig. 1b.
In 1 H magic-angle spinning nuclear magnetic resonance (MAS-NMR) spectra of the crushed AAO samples, a broad chemical shi is observed around 5 ppm, Fig. 1c. This shi can be tted by three Gaussian components corresponding to the protons in physically adsorbed water (d ¼ 5.9 ppm, green dotted line) and to the -OH groups bound to aluminum (d ¼ 3.5 ppm and d ¼ 1.3 ppm, purple and blue dotted line), giving a total number of 2.4 Â 10 20 -OH per g À1 . The total density of the -OH groups is calculated to be 6 nm À2 , taking into account the total number of -OH groups from 1 H MAS-NMR and the calculated surface area, A calc . The predominant adsorption of Tf À anion on the AAO surface in contact with 1 M LiTf/triglyme is expected to lead to an interfacial (or "soggy sand") effect. 33 Indeed, the effective surface zeta potential (z eff ) of crushed 1 vol% AAO particles dispersed in 1 M LiTf/triglyme electrolyte is negative, z eff ¼ À28 mV (z eff ¼ À6 mV in 0.2 LiTf/triglyme). However, when the activity of the -OH groups (surface deprotonation) of crushed AAO membranes was probed by potentiometric titrations, values of IEC ¼ 7.2 Â 10 À2 mmol g À1 are measured, an order of magnitude lower than what has been observed for porous SiO 2 particles in a similar solvent. 38 With this in mind, the interfacial effect in AAO: glyme electrolytes is expected to be milder than in the SiO 2 : glyme-based lithium electrolytes. The AAO: (1 M LiTf/triglyme) electrolyte appears transparent (Fig. S1 †) despite its relatively high volume fraction of Al 2 O 3 solid, 4 ¼ 0.24, and has thermal properties which are comparable to the pure 1 M LiTf/triglyme.
The lithium transference number of AAO: (LiTf/triglyme) was measured using the galvanostatic polarization method as reported previously. 34 In such a measurement, the salt concentration gradient establishes eventually upon applying constant current and at the steady state, the voltage response results solely from the cation (lithium) transport. The contributions of other resistances to the overall voltage increase, such as the resistances of the passivation layers (solid electrolyte interphase, SEI) formed on the lithium electrodes, should also be considered. The lithium transference number can subsequently be calculated from: where I is the constant current applied, R tot,0 is the total cell resistance before the galvanostatic polarization, U N is the steady state voltage and R SEI,0 , R SEI,N are the SEI resistances before and aer the polarization, respectively. As seen in the Fig. 2a, galvanostatic polarization of AAO: (1 M LiTf/triglyme) electrolyte is a fast process, reaching the steady state aer just 200 s, a consequence of the comparatively thin electrolyte. As predicted from the AAO morphology, surface area and the quantity of active -OH groups, lithium transference number calculated using eqn (1) results in t Li ¼ 0.48, considerably higher value than the transference number of the pure liquid electrolyte, t Li ¼ 0.15, at same salt concentration (Fig. S4 †). The semicircle observed in Fig. 2b corresponds to the SEI formed on the lithium metal surface in contact with the liquid part of the electrolyte. Such SEI appears to be stable during the polarization process with approximate thickness around 10 nm, when calculated for the approximate dielectric constant 3 z 10 (calculation available in ESI †). Thus, even though the t Li values reported are of the same order as previously seen for liquid lithium electrolytes, the composite appears more promising due to its stability vs. lithium. Furthermore, the U(t) curve exhibits a clear exponential behavior for longer times, enabling us to determine the LiTf salt diffusion coefficient from (2) l being the electrolyte thickness (here corresponding to the thickness of AAO), and s d being the time constant of the stoichiometry polarization. The time constant is determined from a linear t of the ln|U À U N | vs. t dependence, as seen in Fig. 2c. The corresponding D d ¼ 4 Â 10 À8 cm 2 s À1 is consistent with the values reported previously in a composite mesoporous SiO 2 : (1 M LiTf/diglyme) liquid/solid electrolyte. 34 The polarization experiments are reproducible, giving analogous values for lithium transference number, room temperature ionic conductivity and salt diffusion coefficient (ESI †). Additionally, an enhancement of t Li from the very low values of 0.006 to 0.03 has been observed in the AAO: 1-methyl-1-propylpiperidinium bis(triuoromethylsulfonyl)imide ionic liquid containing 0.2 M LiTf (Fig. S6 †). Although the improvement of t Li in the solid-liquid composite electrolytes in the current case is still lower than observed in the mesoporous SiO 2 : (1 M LiTf/diglyme) due to the lower surface area and lower IEC of AAO, the reproducibility of pore structure provide means to systematically investigate the inuence of other parameters on the lithium transference number and ionic conductivity, including concentration and temperature. Fig. 3a shows ionic conductivity and lithium transference number in AAO: (1 M LiTf/triglyme) electrolyte in the temperature range from 10 to 80 C, where the conductivity of the composite liquid/solid electrolyte improves from 1.8 to 2.6 mS cm À1 . The ratio of the ionic conductivity of the liquid/solid composite and the bulk liquid electrolyte stays constant in the entire temperature range (s m /s N z 0.5), indicating similar conduction mechanisms and partial blocking of some pathways in the liquid/solid composite electrolyte. Indeed, it is even visible from the SEM that not all of the pores in AAO are continuous and interconnected. A more drastic effect has been observed for the lithium transference number of the liquid/ solid composite, leading to the highest value of t Li ¼ 0.80 at 80 C. This can be explained either by activation of additional surface -OH groups of AAO and/or through changes in the molecular speciation in the liquid electrolyte upon heating, investigated using temperature controlled IR spectroscopy (Fig. S2 †). Indeed, temperature increase enhances the ion-pair association to ion pairs and dimers in 1 M LiTf/triglyme (similarly as in LiTf/diglyme 38,39 ). The strong solvation of free ions (about 3 glyme molecules) causes endothermic association reaction and entropy gain through association. The result is a stronger interfacial effect and high t Li in AAO: (1 M LiTf/ triglyme). At higher temperatures, the increase in ionic conductivity accompanied with the depression in concentration of free ions implies a total mobility increase about 4 times. This is in line with the observed viscosity increase in liquid 1 M LiTf/ triglyme (Fig. S3 †).
With increasing temperature, a substantial depletion of the resistance of the SEI layer, R SEI , is observed, Fig. 3b, suggesting its chemical composition or morphological changes. This specic point will be discussed in the subsequent work. Following the mobility trend, salt diffusion coefficient, D d , increases an order of magnitude in the chosen temperature range (Fig. 3c). Fig. 4a shows the effect of salt concentration on room temperature ionic conductivity and lithium transference number in AAO: LiTf/triglyme. A sharp rise (about 2 orders of magnitude) in the ionic conductivity of liquid/solid composite electrolyte upon increasing concentration from 0.02 M to 0.4 M is observed, followed by a plateau and a drop at high salt concentrations.
To discuss the interfacial effect in the present system, two extreme cases of strong and weak anion adsorption can be considered. In the case of weak anion adsorption, space charge effects are negligible andif all pathways percolatingthe conductivity of the composite is simply given by where s N is the bulk liquid conductivity. The slope of the concentration dependent conductivity and the lithium transference number remains identical to the liquid system. For AAO: LiTf/triglyme, a conductivity of 76% of the bulk liquid is expected (24% depression of the bulk).
In the case of strong anion adsorption on AAO, for sufficiently high surface area, we can assume that all the anions are adsorbed, including the ones in the ion pairs. The cations are then all in the space charge zones. As we can neglect bulk contribution, the conductivity of the composite can be written as where u + is the mobility of lithium ions in the pure liquid electrolyte. Here we assume that the mobility of free lithium ions in the space charge zone (double layer) is equal to the mobility of free lithium in the bulk liquid electrolyte. On the other hand, the conductivity of the bulk liquid can be written as The values of c + can be taken from the previous measurements of the triate stretching band by infrared spectroscopy, while the mobilities can be taken from the deconvoluted contributions of specic ions to the overall conductivity (s + , s À ) leading to values of s m /s N z 0.15 at 1 M LiTf and s m /s N z 0.5 at 0.2 M LiTf. 38 In the case of strong adsorption, transference number of 1 is expected.
As the conductivity and lithium transference number are in between the cases of weak and strong adsorption, medium level of adsorption appears to be relevant. A more precise approach would require more information than available. The ratio of the ionic conductivity of the bulk and the ionic conductivity of the liquid/solid composite varies from of s m /s N ¼ 0.8 for 0.02 M LiTf to s m /s N ¼ 0.3 at 1 M LiTf, suggesting stronger adsorption at low c salt and milder adsorption at high c salt . Following the weakening of the adsorption trend, the lithium transference number signicantly decreases upon salt addition from t Li ¼ 0.85 at 0.02 M to t Li ¼ 0.52 at 0.4 M, and the trend continues with somewhat lower rate at higher salt concentrations.
At lower salt concentrations, an additional semicircle corresponding to the electrolyte bulk appears (Fig. 4b). On the other hand, the SEI thickness rises to d z 40 nm. Interestingly, the DSC spectra of AAO: 0.02 M LiTf/triglyme is drastically different to other liquid-solid electrolytes with salt concentration between 0.2-1 M (Fig. 4c) and is almost identical to the DSC of pure triglyme.
To further investigate the cyclability of this class of electrolytes in the symmetric lithium cells, a cyclic lithium platingstripping procedure consisting of one hour cell charging followed by one hour discharging was employed. Fig. 5 illustrates the time-dependent voltage prole under constant current density of 0.3 mA cm À2 for a Whatman glass ber separator inltrated with 1 M LiTf/triglyme, and AAO: 0.02 M LiTf/ triglyme electrolyte, respectively. The voltage response of a Whatman glass ber separator inltrated with 1 M LiTf/ triglyme increases linearly in time to values as high as 0.5 V aer 300 hours. In contrast, the AAO: 1 M LiTf/triglyme shows no gradual voltage increase, and voltage stabilizes with time achieving values of 0.1 V at the end of the stripping-platting procedure (see inset). However, sudden and momentary instabilities can be observed in the case of AAO: 1 M LiTf/triglyme. This may be consequence of the blocking of some pores and appearance of local short circuits due to the electrolyte and/or salt decomposition. 40 Such behavior has been observed for both liquid electrolytes inltrated in separator as well as in solid electrolytes with highly stable SEIs and does not interfere with long-term electrolyte performance. 41 The cell failure in the AAO: 1 M LiTf/triglyme electrolyte seems to occur later than its AAOfree counterpart, suggesting that liquid-solid are able to partially hinder dendrite formation and substantially enhance the lifetime of the cell. On the other hand, one might expect that the AAO: 0.02 M LiTf/triglyme would show the best stripping/ plating behavior since there the lithium transference number is the highest. This has however not been experimentally conrmed and is most probably the consequence of unstable SEI in the electrolytes with low salt concentration, inducing dendrite formation. 42 Growth of lithium deposits in the form of mossy lithium can be observed by ex situ SEM aer disassembling Li|AAO: (1 M LiTf/triglyme)|Li cell both on the surface of lithium and on the AAO (Fig. 6a-c). The deposits are large particles (ca. 100 mm long), covered by a thin SEI layer. This is conrmed by energydispersive X-ray spectroscopy (EDS) elemental mapping images (Fig. 6c) showing that C, F, and O elements are evenly distributed on the particle surface. The structure of the deposit appears to be crystalline ( Fig. 6d and e), with clear lattice fringes visible in high resolution TEM. However, the amount of the crystalline phase is not large enough to be detected by the XRD measurements. Such deposits might lead to cell short circuit if they penetrate the full length of AAO.

Conclusions
In this study, we report on LiTf/triglyme inltrated in porous AAO as a possible electrolyte for lithium metal batteries. The benecial adsorption of the anion on the surface of AAO leads to the increase of lithium transference number, t Li . The interfacial effect on the surface is higher at lower salt concentration and elevated temperature, with t Li reaching values of 80% at 80 C in 1 M LiTf/triglyme, and 85% at room temperature in 0.02 M LiTf/ triglyme. Interestingly, glyme-based AAO electrolyte seems to combine high lithium transference numbers with good stability with lithium electrodes.
Glyme-based liquid-solid electrolytes proved to be more successful in stopping dendrite growth compared to their liquid counterparts inltrated in glass ber separators. However, at long stripping/plating times, mossy crystalline lithium deposits covered by an SEI can still be observed.

Conflicts of interest
There are no conicts to declare.