An alternative route to single ion conductivity using multi-ionic salts

Abstract

Multi-ionic lithium salts comprised of polyoligomeric silsesquioxanes (POSS) functionalized with eight – (LiNSO₂CF₃) groups, referred to as POSS-(LiNSO₂CF₃)₈, can be dissolved at very high loadings into tetraglyme (G₄), where they can be considered solvent-in-salt electrolytes. With increasing dilution, colloidal solutions are formed. Two systems were investigated, neat POSS-(LiNSO₂CF₃)₈ in G₄ and mixtures of POSS-(LiNSO₂CF₃)₈ with LiPF₆ or lithium bis(trifluoromethanesulfonyl)imide (LiTFSI). PFG-NMR indicates that Li can be un-dissociated, completely dissociated and surrounded by G₄ molecules, or as contact ion pairs (in which there are 3–4 ether oxygen contacts and one contact with the oxygen from the anion). Equilibria exist between the species of POSS-(LiNSO₂CF₃)₈ and if there is rapid equilibration between the Li states, and close enough proximity between the POSS-(LiNSO₂CF₃), then the Li⁺ ions can migrate by a Grotthus-type coordinated hopping mechanism, as well as by a purely diffusive motion. Unlike polymer single ion conductors, where the backbone flexibility permits cluster/aggregate formation, which inhibits escape and mobility of the Li⁺ ions, the rigid POSS cube and its colloidal structure in G₄ prevents formation of POSS-(NSO₂CF₃⁻)⋯Li⁺⋯(⁻CF₃NSO₂)-POSS triplets. Instead, the solvated Li⁺ in POSS-(NSO₂CF₃⁻)⋯Li⁺⋯G₄ can be more easily removed to form conductive G₄⋯Li⁺⁻⋯G₄. Good ionic conductivities (∼10⁻⁴ S cm⁻¹) and lithium ion transference numbers of t^PP₊ = 0.65 can be achieved in these systems. Mixtures of 80 wt% LiTFSI and 20 wt% POSS-(LiNSO₂CF₃)₈ in G₄ at an O/Li ratio of 20/1, yield both high conductivity (σ = 3.3 × 10⁻³ S cm⁻¹) and high (t^PP₊ = 0.65) transference number. Stable cycling between C/2 and 2C with high capacity retention was achieved using Li/[G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈]/LiFePO₄ half-cells.

---

Conceptual insights

Critical goals for the achievement of safe, nonvolatile solid organic electrolytes in lithium ion batteries are the development of materials with both high ionic conductivity (σ) and lithium ion transference numbers, t_Li⁺. Liquids offer advantages over solids because they can permeate porous cathodes and access the active components more effectively. This article shows that the use of multi-ionic lithium salts based on polyoligomeric silsesquioxanes (POSS) with pendant anions, POSS-(LiNSO₂CF₃)₈, can result in both high ionic conductivity (σ > 10⁻⁴ S cm⁻¹) and high lithium ion transference numbers (t_Li⁺ = 0.65) when dissolved in low volatility tetraglyme (G₄) solvents. Solvent in salt electrolytes are formed at high POSS-(LiNSO₂CF₃)₈ concentrations and colloids are formed with increasing dilution. The absence of aggregation in the partially charged, rigid, symmetric POSS-(LiNSO₂CF₃)₈ in tetraglyme prevents trapping of Li⁺ ions that can slow down Li⁺ ion migration in polymer single ion conductors. Li⁺ ion migration can occur by both diffusive and coordinated hopping mechanisms. The addition of mono-ionic lithium salts increases conductivity (σ > 10⁻³ S cm⁻¹) while maintaining t_Li⁺ = 0.65.

Introduction

One of the major problems in the development of separators for lithium-ion or metal batteries (LIBs, LMBs) is to optimize ionic conductivity (σ) and reduce concentration polarization. Concentration gradients adversely affect cell performance by contributing to the growth of dendrites, which are branched or needle-like structures that form instead of desirable flat, uniform Li⁰ deposition. Dendrites can break off, resulting in a decrease in energy density, or span the cell separator causing internal shorting, heating, thermal runaway and catastrophic cell failure. Dendrites are particularly a concern for lithium metal batteries (LMBs), but also occur in lithium-ion batteries (LIBs), since Li⁰ metal rather than intercalated Li⁺ can deposit on the anode in LIBs during fast charging or at low temperatures.¹ The transition from graphitic to metallic lithium anodes would enhance energy density 10 fold, and metallic lithium anodes would enable the use of un-lithiated materials such as sulfur or air to replace intercalation cathodes for lithium/sulfur or lithium/air batteries^2,3 with improved energy density. In these cases, the safety issues associated with dendrite growth when using metallic Li⁰ are a major concern.

There has been extensive theoretical and experimental research on the prevention of dendritic growth, including mechanical inhibition⁴ and limiting concentration gradients that result in anion depletion near the anode.⁵ The latter can be avoided by the use of electrolytes with high ionic conductivities and low anion mobility, i.e. high Li⁺ ion high transference numbers (t_Li⁺). The transference number is by definition the contribution of a single charged species to transport under the influence of an electric field. Since t_Li⁺ + t_X⁻ = 1, if the anion is immobile t_X⁻ = 0, t_Li⁺ = 1, and concentration polarization does not develop. A t_Li⁺ = 1 was shown to enhance power and energy density, particularly at high discharge rates, over electrolytes with t_Li⁺ = 0.2, even with an order of magnitude decrease in conductivity.⁶ Recent modelling studies suggest that t_Li⁺ ≈ 0.7 would also allow a higher attainable state of charge at high charge rates.⁷

However, most electrolytes are not ideal and lithium can be transported in the form of neutral contact ion pairs, and neutral or charged aggregates. When the electrolyte is non-ideal, an equilibrium exists between Li⁺, X⁻, LiX, Li₂X⁺, LiX₂⁻ (and even larger aggregates), and t_Li⁺ depends on the concentration (activity), mobility and charge of each charged species.⁸ The measurement of t_Li⁺ is therefore not straightforward and an accurate determination is not always possible.⁹ The two most common methods used to measure transport properties of lithium are spin echo pulse-field gradient (PFG)-NMR and electrochemical techniques,¹⁰ particularly potentiostatic polarization, developed by Bruce and Vincent.^11,12 In spin echo PFG-NMR,^13–16 the average static self-diffusion coefficients of all species containing a nucleus with a unique chemical shift is monitored, whereas only charged species are measured using potentiostatic polarization. Both methods can overestimate the effective t_Li⁺, and they often do not agree with each other.

Concentration gradients are avoided in polymer single ion conductors (SICs) by covalent attachment of anions to the polymer backbone so that t_Li⁺ → 1. However, conductivities of polymer SICs have remained low (σ < 10⁻⁵ S cm⁻¹).¹⁷ In polymer electrolytes the cation motion is believed to be coupled to the backbone dynamics, so attempts to increase conductivity have included incorporating flexible chains with low glass transition temperatures (T_gs) such as polydimethylsiloxanes. One of the reasons for the low conductivity in polymer SICs is extensive ion aggregation. As in the case of bi-ionic conductors, there are completely dissociated ions, neutral contact ion pairs (CIPs) and poorly conductive ion triplets and aggregates (AGG) (Fig. 1A). However, for polymer SICs larger ion aggregates also form, as shown schematically in Fig. 1B and there are few dissociated ions.^18,19 In this case, polymer flexibility adversely affects conductivity by contributing to the ion clustering, since the pendant ion groups have the mobility necessary to phase separate from the more hydrophobic backbone. Confirmation that the formation of these aggregates causes a decrease in ionic conductivity comes from the unexpected finding that addition of a tetraglyme plasticizer to an ionomer resulted in a reduction of ionic conductivity by 4–5 orders of magnitude, as the result of recoupling of the ionic conductivity with the polymer segmental dynamics.²⁰ Single ion conductors can be crosslinked and swollen with liquid solvents to achieve t_Li⁺ = 1 and higher ionic conductivity than dry systems^21,22,23 and as in the dry systems, highly dissociative ionic species reduce the formation of aggregates and increase conductivity. Thus, replacement of carboxylate, phosphate and sulfonate anions with the LiTFSI-like anion in lithium[(4-styrenesulfonyl)trifluoromethane-sulfonyl)imide]^24,25 and poly[(4-styrenesulfonyl)(trifluoromethyl(S-trifluoromethylsulfonylimino)sulfonyl)imide]²⁶ have resulted in the best ion conductivities to date.

Fig. 1 Schematic of (A) ion pairs and ion triplets found in LiX salts and polymer SICs; (B) higher order aggregates that can occur in polymer single ion conductors (SICs); (C) solvate ionic liquid [Li(G₄)⁺]/[TFSI⁻], mixtures of G₄ and LiTFSI and mixtures of G₄ and POSS-(NSO₂CF₃⁻)Li⁺; (D) structure of POSS-(LiNSO₂CF₃)₈; and (E) POSS-(NSO₂CF₃⁻)₈ anions dispersed in G₄; G₄ not shown.

Nonflammable liquids have some advantages over solids when used as battery electrolytes. In particular, they can permeate the space between the composite electrodes so that more of the active material is accessible. In order to increase t_Li⁺ in liquids, the strategy has been to use large, slow moving anions. Anions tethered to SiO₂ nanoparticles²⁷ and polyanions dissolved in liquid solvents^28–30 have been investigated, and can have high room temperature conductivities (σ > 10⁻⁴ S cm⁻¹) and t_Li⁺ > 0.7. Recently investigated solvent-in-salt electrolytes also exhibit t_Li⁺ ∼ 0.7 and high conductivity.³¹ Tetraglyme (G₄), CH₃–O–(CH₂CH₂O)₄–CH₃, is an alternative to expensive nonflammable ionic liquids (ILs), since it forms a solvate (or chelate) IL, where a third component (here G₄) strongly coordinates Li⁺, thus forming a complex cation. This has been shown to occur for a 1/1 molar mixture of G₄ and LiTFSI, [Li(G₄)⁺][TFSI⁻], with brackets used to indicate the molar ratios, while for other compositions or salts, coordination of Li⁺ can occur between both G₄ and the anion (Fig. 1C).^32–38

Here we propose the use of functionalized symmetric, multi-ionic polyhedral oligomeric silsesquioxane (POSS) with dissociative lithium salts (Fig. 1D) dissolved in tetraglyme (G₄), CH₃–O–(CH₂CH₂O)₄–CH₃, to increase lithium ion transference numbers (t_Li⁺) and avoid the formation of large ion clusters that trap the Li⁺ ions. The salts are composed of (SiO_1.5)₈ cubes, where the eight corners have been functionalized with LiTFSI-type salts (see Experimental details, synthesis and Scheme S1 in ESI†), and will be referred to as POSS-(LiNSO₂CF₃)₈. The POSS-(LiNSO₂CF₃)₈ can be considered solvent-in-salt electrolytes, since salt weights = 3× solvent weight can be achieved.³¹ When dissolved in G₄, some of the Li ions dissociate, resulting in partially charged species, and form a colloidal solution. These large negatively charged anions repel each other, preventing aggregation Fig. 1E. We have investigated POSS-(LiNSO₂CF₃)₈ in G₄, and compared the properties of G₄/POSS-(LiNSO₂CF₃)₈ with those of G₄/LiX, X = TFSI⁻, PF₆⁻, and lithium bis(perfluoroethanesulfonyl)imide (BETI⁻) = [N(SO₂C₂F₅)₂]⁻, and their mixtures, G₄/POSS-(LiNSO₂CF₃)₈/LiX. Mixtures of POSS-(LiNSO₂CF₃)₈ and LiTFSI or LiPF₆ dissolved in G₄ have t_Li⁺ ∼ 0.65 in the liquid state, with conductivities σ > 10⁻³.

Results and discussion

The preparation and compositions of all the samples are given in Table S1 (ESI†). G₄/POSS-(LiNSO₂CF₃)₈ and G₄/LiX, X = TFSI⁻, PF₆⁻ and BETI⁻, were characterized based on the ratio of ether oxygens in G₄ to Li⁺ ions (O/Li ratio), as is common in PEO based electrolytes, and/or their molar ratios. In the case of the 1/1 molar ratio for G₄ and LiX, this is O/Li = 5/1. In the case of G₄/POSS-(LiNSO₂CF₃)₈, a 1/1 molar ratio is O/Li = 5/8, while to provide one G₄ for all of the eight Li on the POSS-(LiNSO₂CF₃)₈, requires O/Li = 40/1. In mixtures of POSS-(LiNSO₂CF₃)₈ with LiX, the weight ratio of POSS-(LiNSO₂CF₃)₈ to LiX varied from 100/0 to 0/100.

Temperature dependent ionic conductivities (see ESI† for experimental details) of G₄/POSS-(LiNSO₂CF₃)₈ and G₄/LiTFSI as a function of O/Li, (Fig. 2 and Table S2, ESI†) show about an order of magnitude decrease in conductivity of G₄/POSS-(LiNSO₂CF₃)₈ compared with that of G₄/LiTFSI. This is encouraging since if the pendant anion were on a polymer SIC, a 2–4 order of magnitude decrease in conductivity is expected. The conductivity of G₄/LiTFSI agrees well with previously reported values at 30 °C (6.56 × 10⁻³ S cm⁻¹).³⁹ Of further interest is the temperature dependent viscosity of both electrolytes (Table S3, ESI†), where the viscosity of the G₄/LiTFSI is only ∼21% less than that of the G₄/POSS-(LiNSO₂CF₃)₈ at 20 °C. This suggests that large scale aggregation does not occur in the G₄/POSS-(LiNSO₂CF₃)₈ (which would otherwise result in large viscosity increases). These results suggest that the difference in conductivity is related to a decreased contribution from the large POSS-(NSO₂CF₃⁻)₈ anions (2520 g mol⁻¹) compared with the TFSI⁻ anion (215 g mol⁻¹) and/or to the better dissociative properties of the LiTFSI (CF₃SO₂NLiSO₂CF₃) compared with –(LiNSO₂CF₃)₈, which has only half the number of electron withdrawing groups. As in the case of bi-ionic conductors and polymer SICs, the conductivity increases as the electron withdrawing groups of the anion increase, and as the negative charge becomes more delocalized. Addition of LiTFSI to G₄/POSS-(LiNSO₂CF₃)₈ increases the conductivity, as expected (Table S2, ESI†).

Fig. 2 Temperature dependent ionic conductivity as a function of O/Li ratio for G₄/LiTFSI and G₄/POSS-(LiNSO₂CF₃)₈.

The absence of aggregation and the formation of a stable colloid for G₄/POSS-(LiNSO₂CF₃)₈ are indicated by the lack of phase separation or precipitation over months, even though the weight of POSS-(LiNSO₂CF₃)₈ is up to 3× greater than the weight of G₄ (see Table S1, ESI†). TEM images (Fig. 3) also show no aggregates, and EDAX data (Fig. S1, ESI†) confirm that these images contain POSS-(LiNSO₂CF₃)₈.

Fig. 3 TEM image of G₄/POSS-(LiNSO₂CF₃)₈ with O/Li = 7.5/1. Sample was prepared by dissolution of POSS-(LiNSO₂CF₃)₈ in polyethylene glycol monomethyl ether acrylate (M_n = 350 g mol⁻¹), 0.1% AIBN, and dilution with acetonitrile (ACN). After dropping the solution on a TEM grid, and evaporation of the ACN, the sample polymerized.

Differential scanning calorimetry (DSC) data (Table S4 and Fig. S2, ESI†) of the neat G₄ indicated it melts at T_m = −28.3 °C in agreement with literature values (T_m = −30³⁵), with no glass transition temperature (T_g). Attempts to crystallize POSS-(LiNSO₂CF₃)₈ were unsuccessful, and it decomposed before T_g was reached. At comparable O/Li ratios, G₄/LiTFSI more effectively suppresses crystallization than does POSS-(LiNSO₂CF₃)₈. For G₄/LiTFSI the melt is completely suppressed for O/Li = 15/1 to 5/1, and there are only very small (crystallization and remelt) peaks for the 20/1 and 17.5/1 samples. For G₄/POSS-(LiNSO₂CF₃)₈, the melt temperature T_m only completely disappear at O/Li = 7.5/1 and 5/1 and there is a very small crystallization and remelt at O/Li = 10/1. Since Li⁺ ions are typically solvated by a combination of 4–5 ether oxygens and/or anion contacts, in the case of G₄/POSS-(LiNSO₂CF₃)₈, more of these contacts may come from contact ion pairs than from the ether oxygens of G₄ than in the case of [Li(G₄)⁺]/[TFSI⁻], where the Li⁺ ions can be completely solvated by G₄ (Fig. 1C). This would leave more G₄ molecules available for crystallization in the case of G₄/POSS-(LiNSO₂CF₃)₈ compared with G₄/LiTFSI, as observed.

When both G₄/POSS-(LiNSO₂CF₃)₈ and G₄/LiTFSI are amorphous, T_g of G₄/LiTFSI is 10 °C higher than G₄/POSS-(LiNSO₂CF₃)₈ at the same O/Li ratio. This also suggests that there is more Li⁺ dissociation in G₄/LiTFSI raising its T_g, while in G₄/POSS-(LiNSO₂CF₃)₈ the Li⁺ ion dissociates less and forms more contact ion pairs (–NSO₂CF₃⁻⋯Li⁺⋯G₄). Both G₄/LiTFSI and G₄/POSS-(LiNSO₂CF₃)₈ show the expected increase in T_g with increasing salt concentration, which can be been attributed to the effects of increased viscosity and –CH₂CH₂O⋯Li⁺⋯OCH₂CH₂– crosslinks with increased salt concentration.

Thermogravimetric analysis (TGA) data (Fig. 4) also provide information on the complex formed between the Li⁺ cation and G₄. Unlike the case of [Li(G₄)⁺][TFSI⁻] (1/1 mole ratio, O/Li = 5/1), where removal of G₄ is delayed until ∼200 °C³⁶ compared with G₄ removal at ∼100 °C in compositions with excess G₄, delayed removal of G₄ does not occur for any composition of G₄/POSS-(LiNSO₂CF₃)₈ (Fig. S3, ESI†) except for the 4/1 mole ratio (O/Li = 2.5/1) (Fig. 4). The TGA data therefore indicates that for G₄/POSS-(LiNSO₂CF₃)₈ (O/Li ≥ 5/1) the Li⁺/G₄ complex is not formed. Instead, the Li⁺ ion is more loosely complexed with the G₄, and possibly also interacting with the NSO₂CF₃⁻ anions of POSS-(NSO₂CF₃⁻)₈. For G₄/POSS-(LiNSO₂CF₃)₈ (4/1 mole ratio, O/Li = 2.5/1), G₄ comes off predominantly at 300 °C, at an even higher temperature than the pure POSS-(LiNSO₂CF₃)₈. In this case, the G₄ is even more tightly complexed than for the [Li(G₄)⁺][TFSI⁻], which is removed at 200 °C. This arises since one G₄ must be shared between two Li⁺ of G₄/POSS-(LiNSO₂CF₃)₈ so that the G₄ is trapped between the cubes, as shown in a 2D representation (Fig. 4). It should be noted that even for this composition, where the weight of the POSS-(LiNSO₂CF₃)₈ is 3× that of the G₄, the mixture is stable against any signs of phase separation or precipitation for months.

Fig. 4 TGA weight loss data: The samples were heated to 100 °C (at 10 °C min⁻¹), held at that temperature for 1 h, and then heated to 400 °C at 10 °C min⁻¹.

Structure of the liquids by X-ray scattering

Wide angle X-ray scattering (WAXS) data obtained at room temperature for neat G₄ and neat POSS-(LiNSO₂CF₃)₈ are shown in Fig. 5A. G₄ crystallized at low temperatures (T_m = −30 °C; shown at −173 °C), but POSS-(LiNSO₂CF₃)₈, which is a glass, had similar WAXS at all temperatures (not shown). At room temperature, G₄ shows a single broad peak centered at 2θ = 21–22° (q ∼ 15 nm⁻¹; d ∼ 0.45 nm) and is a feature typical of molecular liquids that can distinguish between intra- and intermolecular distances.⁴⁰ POSS-(LiNSO₂CF₃)₈ exhibits a shoulder peak at 2θ ∼ 22° (q ∼ 15.5 nm⁻¹; d ∼ 0.40 nm), and two broad features at 2θ ∼ 18.5° (q ∼ 13.1 nm⁻¹; 0.48 nm) and 2θ ∼ 6.9° (q = 4.9 nm⁻¹, d = 1.28 nm). Assignments of the peaks can be made in analogy with LiTFSI in the (simulated) liquid state (530 K). For LiTFSI, deconvolution of S(q) obtained from MD simulations⁴¹ in the 4 < q/nm⁻¹ < 20 region show three Gaussian functions: a shoulder peak at 2θ = 21.94° (q ∼ 15.5 nm⁻¹; d = 0.405 nm), and peaks centered at 2θ = 16.8° (q = 11.9 nm⁻¹; d = 0.53 nm) and 2θ = 8.7° (q = 6.2 nm⁻¹; d = 1.01 nm) and. The 2θ = 21.94° (q ∼ 15.5 nm⁻¹; d = 0.405 nm) shoulder peak in LiTFSI, known as the contact peak, corresponds to distances between neighboring atoms of different ions, defining the boundary between intra- and inter-molecular features,⁴¹ and can be similarly assigned for amorphous POSS-(LiNSO₂CF₃)₈. The 2θ = 16.8° feature in LiTFSI has been attributed to a charge-ordering peak (COP), often referred to as the polar network, and is a feature observed in ionic liquids and molten salts, since ions need to be surrounded by counter-ions to establish local electroneutrality.⁴² In LiTFSI, this assignment is corroborated by a more ordered COP at d = 0.52 nm in the crystalline state.⁴³ The 2θ =18.5° (q ∼ 13.1 nm⁻¹; d = 0.48 nm) feature in POSS-(LiNSO₂CF₃)₈ can also possibly be assigned to the COP peak, but crystalline POSS-(LiNSO₂CF₃)₈ is not available to make a similar comparison. Finally, the 2θ = 8.7° (6.2 nm⁻¹, 1.01 nm) peak in LiTFSI was attributed to distances between sulfur atoms belonging to two TFSI⁻ ions, oriented side by side without an intervening Li, and occurs at 0.86 nm in crystalline LiTFSI.⁴² By analogy, the 2θ = 6.9° (q ∼ 4.9 nm⁻¹; 1.28 nm) peak in POSS-(LiNSO₂CF₃)₈ may correspond to some larger distance between the sulfur containing pendant anions. Small angle X-ray scattering (SAXS) data (not shown) showed no structure at small q for either G₄ or POSS-(LiNSO₂CF₃)₈.

Fig. 5 (A) WAXS data for neat G₄ (25 and −173 °C), neat POSS-(LiNSO₂CF₃)₈, and mole ratios of [G₄]/[POSS-(LiNSO₂CF₃)₈] also shown as (O/Li ratio); and (B) SAXS data at 25 °C for [G₄]/[POSS-(LiNSO₂CF₃)₈], O/Li = 7.5/1 obtained in 2 regions (, ); and the (C) S(q) and (D) P(R) data.

WAXS data for mixtures of G₄/POSS-(LiNSO₂CF₃)₈ at room temperature with O/Li = 2.5/1, 5/1, 7.5/1 are shown in Fig. 5A. Temperature dependent data (not shown) were identical, and none of the samples showed crystalline peaks of G₄ at −100 °C. With the addition of G₄ to POSS-(LiNSO₂CF₃)₈ the (shoulder) contact peak of POSS-(LiNSO₂CF₃)₈ at 2θ ∼ 22° (q ∼ 15.5 nm⁻¹; 0.40 nm) shifts to lower angles, i.e. larger distances, and merges with that of G₄, and the intense charge ordering peak at 2θ ∼ 18.5° (q ∼ 13.1 nm⁻¹; 0.48 nm) disappears. In the case of G₄/LiTFSI (1/1 molar ratio), the contact peak (shoulder) at 2θ = 21.94° (q ∼ 15.5 nm⁻¹; d = 0.405 nm) also shifts to lower values of 2θ = 21.2° (q ∼ 15.0 nm⁻¹; d = 0.418 nm), i.e. larger distances, and the intense charge ordering peak also disappears. The addition of G₄ to POSS-(LiNSO₂CF₃)₈ for O/Li = 2.5/1 and 5/1 results in a shift of the feature at 2θ ∼ 6.9° (q = 4.9 nm⁻¹, d = 1.28 nm), tentatively assigned to separation between the anionic side-groups, to lower angles 2θ = 5.48° (q = 3.9 nm⁻¹; d = 1.61 nm), consistent with an increase in the separation distance between POSS-(LiNSO₂CF₃)₈ due to intervening G₄ molecules. In the case of G₄/LiTFSI a new peak appears that shifts from higher angles to 2θ = 13.38° (q ∼ 9.5 nm⁻¹, d = 0.66 nm) for the 1/1 molar ratio, and is attributed to a characteristic separation distance between anions consistent with glyme-solvated Li ions.^40,41

The peak associated with anion separation in G₄/POSS-(LiNSO₂CF₃)₈ at 2θ ∼ 6.9° (q = 4.9 nm⁻¹, d = 1.28 nm) can only be observed for G₄/POSS-(LiNSO₂CF₃)₈ with O/Li = 2.5/1 and 5/1, which correspond to the mole ratios [4G₄][1 POSS-(LiNSO₂CF₃)₈] and [8G₄][1 POSS-(LiNSO₂CF₃)₈], where 1 G₄ is shared between two Li or there is 1 G₄ for each Li, respectively. Since it was not possible to investigate values of 2θ < 5° by WAXS (due to instrumental limitations), SAXS was used to determine whether aggregation/clustering or ordering could be observed for G₄/POSS-(LiNSO₂CF₃)₈ for O/Li = 7.5/1. This sample was chosen since it had the highest concentration of POSS-(LiNSO₂CF₃)₈ whose viscosity allowed it to be put into narrow quartz capillaries. SAXS data were obtained in two q regions, which were spliced together (Fig. 5B). Subtraction of G₄ from G₄/POSS-(LiNSO₂CF₃)₈ in the 3 to 20 q (nm⁻¹) region showed that the 2θ = 18.5 peak of POSS-(LiNSO₂CF₃)₈ persisted in the mixture and was distinct from the close 2θ = 21–22 peak (q ∼ 15 nm⁻¹; d = 0.45 nm) of G₄; subtraction of G₄ was not done in WAXs data, Fig. 5A. SAXS data from q = 1 nm⁻¹ to q = 5 nm⁻¹ obtained for G₄/POSS-(LiNSO₂CF₃)₈ with O/Li = 7.5/1 was fit to a structure factor, assuming a model where the primary particles were hard spheres (Fig. 5C). The form factor (Fig. 5D) gave “particles” of 1.57 nm radius. The SiO_1.5 core has an approximate radius of 0.45 nm.⁴⁴ The large radius observed by SAXS suggests that the POSS-(LiNSO₂CF₃)₈ cubes do not aggregate and instead form solvated structures, with the space between adjacent SiO_1.5 cores of POSS-(LiNSO₂CF₃)₈ occupied by free G₄, G₄ participating in contact ion pairs (one anion –NSO₂CF₃⁻⋯Li⁺ contact and 3–4 ether oxygen –Li⁺ contacts from G₄), and solvated Li⁺G₄. If the “particles” were in a cubic array, the center to center distance would be approximately 1.64 nm apart, smaller than the center to center distance of the “particles” measured by SAXS. However, it is likely that the structure factor should consider interpenetrating particles.

Lithium ion transference numbers (t^PP₊) by potentiostatic polarization

Lithium ion transference numbers (Table 1) were obtained by potentiostatic polarization experiments (t^PP₊), where they were calculated (eqn (1)) after corrections for the solid electrolyte interphase (SEI) that forms on the electrodes (eqn (2)).Here ΔV is the applied small constant potential (ΔV), with the electrolyte between Li⁰ metal non-blocking electrodes, I_total is the initial current (I₀), composed of both anions and Li⁺ cations, I_Li⁺ is the steady state current (I_ss) composed of only Li⁺ ions, D₊ and D₋ are the diffusion coefficients of the cation and anion, respectively, and R₀ and R_ss are the initial and final interfacial resistances. Since we measured I_initial at t = 1 s, as recommended,⁴⁵ not observing the initial current drop is probably not a source of error in t^PP₊.

Table 1 Diffusion coefficients, lithium ion transference numbers, and ionic radii (or normalized ionic radii) for G₄/LiPF₆, G₄/POSS-(LiNSO₂CF₃) and mixtures with O/Li = 20/1

Sample O/Li = 20/1	Diffusion constant, D, m^2 s^−1 × 10^11				Lithium ion transference numbers			Radii (nm) or normalized radii
								R ^is from η^b
	D G4	D PF6	D 8mer	D Li	t ^NMR+	t ^PP+	t ^NMR−	R ^Lis		R ^POSSs	Li	PF6	POSS
a 8mer = POSS-(LiNSO2CF3). b (@25 °C, where η (mPa s), D (m^2 s^−1 × 10^−11). c If equal moles of two species are assumed. d O/Li = 18.3/1.
G4/LiPF6	7.70	5.83		4.41	0.43	0.51	0.57	0.33	0.25		1.75	1.32
G4/(LiPF6 + 20 wt% 8mer^a)^d	6.41	4.97	1.14	3.53	0.42	0.65	PF6^− 0.583	0.33	0.24	1.03	1.81	1.29	5.6
							POSS^− 0.002
G4/8mer^a	11.3		2.05	1.6	0.44	0.65	0.60^c	0.92		0.63	7.1		5.5
				1.1	0.35			1.34			10.3

---

What is interesting is that for all the G₄/LiX (X = TFSI⁻, PF₆⁻ and BETI⁻), t^PP₊ = 0.5 ± 0.01, but for G₄/POSS-(LiNSO₂CF₃)₈, and all the mixtures, i.e. G₄/LiX/POSS-(LiNSO₂CF₃)₈, t^PP₊ = 0.65 ± 0.01. Any system with greater than 10 wt% POSS-(LiNSO₂CF₃)₈ also has t^PP₊ > 0.65 (Table S2, ESI†).

Evidence that we do not have significant amounts of neutral species (although we could have other charged species) of POSS-(LiNSO₂CF₃)₈ or LiX in G₄ comes from measurements of σ_eff (σ_eff = I_ss/ΔV, σ₀ = I₀/ΔV). σ_eff is expected to be independent of the applied voltage when ΔV: (i) ≤10 mV for a fully dissociated ideal electrolyte;¹¹ (ii) <10 V for high concentrations of LiX (since the solution acts like an insulator).⁴⁶ Therefore measurement of the steady state current itself can indicate when significant ion pairs are present.⁴⁷ For all of the systems we investigated, values of ΔV = 10, 20 and 30 mV were tested, and the current increased as ΔV increased, indicating the absence of significant neutral species.

Although measurements of σ_eff/σ₀ (σ_eff = I_ss/ΔV, σ₀ = I₀/ΔV) do not yield t₊ (eqn (1)) when there are ion pairs or multiply charged species (since t₊ is defined for a single species, typically Li⁺), it is important to stress that in practical applications, what is of equal importance is the steady state DC current, which is a measure of the net flux of the electroactive constituent (lithium) in the cell. Although the initial DC current (I₀) under the same applied potential (∼20 mV) is about 3 times higher for LiX (X = TFSI⁻, PF₆⁻, BETI⁻) than for POSS-(LiNSO₂CF₃)₈, the steady state current (I_ss) is always higher for the POSS-(LiNSO₂CF₃)₈ or mixtures of POSS-(LiNSO₂CF₃)₈/LiX than for LiX. A high value of σ_eff is desirable, since in battery applications, during discharge, lithium can be carried by Li⁺, aggregates and LiX. It is less important which species carries the current, provided that the total flux of Li from anode to cathode is high, which is the case here.

Diffusion measurements and lithium ion transference (t^NMR₊) from PFG-NMR

Diffusion coefficients obtained by PFG-NMR are summarized in Table 1. In order to also measure the mixed diffusion of POSS-(LiNSO₂CF₃)₈ and LiX in G₄, LiPF₆ was selected since it has a distinct ¹⁹F NMR signal compared with –CF₃ from POSS-(LiNSO₂CF₃)₈. Three systems were investigated: G₄/POSS-(LiNSO₂CF₃)₈, G₄/LiPF₆ and G₄/[80 wt% LiPF₆ + 20 wt% POSS-(LiNSO₂CF₃)₈].

¹H-NMR (Fig. S5, ESI†) was only used to measure the diffusion coefficient of the G₄, using chemical shifts in the 3.5–5 ppm range. The diffusivity of the PF₆⁻ was measured by ¹⁹F PFG-NMR at its characteristic signal, a doublet resonating at around −72 ppm, and the diffusivity of the POSS anion was followed using the CF₃ resonance at −76 ppm (Fig. 6). As expected, the mixed G₄/(LiPF₆ + 20 wt% POSS-(LiNSO₂CF₃)₈) samples had features from both the PF₆⁻ and the –CF₃ species. Of interest is that in the ⁷Li spectra of neat G₄/POSS-(LiNSO₂CF₃)₈, two broader but separate peaks are observed.

Fig. 6 ⁷Li and ¹⁹F PFG-NMR data for G₄/LiPF₆, G₄/POSS-(LiNSO₂CF₃)₈, and G₄/(LiPF₆+ POSS-(LiNSO₂CF₃)₈).

In all cases, the self-diffusion constants increase in the order D_G4 > D_anion > D_Li⁺, the same order observed previously for dilute LiX solutions of glyme/LiX^36,39 and other aprotic solvents. The diffusion constant of neat G₄ decreased with the addition of G₄/LiPF₆, G₄/POSS-(LiNSO₂CF₃)₈, and G₄/(LiPF₆+ POSS-(LiNSO₂CF₃)₈), consistent with previous reports for addition of LiX salts to glymes.^36,39 Since D_G4 has the fastest diffusion constant, it indicates that there is free glyme in all the solutions. Further, D_G4 is fastest for G₄/POSS-(LiNSO₂CF₃)₈, indicating that there is more free G₄ and less G₄ participating in solvation shells than for G₄/LiPF₆, consistent with the DSC results. The diffusion constants of Li⁺ and PF₆⁻ in G₄/LiPF₆ are slower by a factor of ∼2 compared with Li⁺ and TFSI⁻ in G₄/LiTFSI (O/Li = 20/1) (D_TFSI⁻ = 11 × 10⁻⁷, and D_Li⁺ = 9 × 10⁻⁷ cm² s⁻¹) at 30 °C³⁶) consistent with the known low lattice energy of LiTFSI compared with other LiX salts (such as LiPF₆) and its tendency for less ion pair formation in solution.

In the case of POSS-(LiNSO₂CF₃)₈, the most salient feature is that for G₄/POSS-(LiNSO₂CF₃)₈, the anion shows a single peak in the ¹⁹F NMR spectra, while Li exhibits two peaks in the ⁷Li NMR spectra. For ¹⁹F, this may indicate a single diffusing species or a variety of diffusing species with rapid chemical exchange between them. However, the existence of 2 peaks in the ⁷Li NMR spectrum indicates that there are two lithium “pools”, and that each lithium pool consists of lithium containing species that exchange with each other and/or have similar chemical shifts. The broadness of the ⁷Li lines is indicative of a lower mobility for the Li⁺ ions.

While we cannot definitely assign the two Li⁺ peaks, the narrower (“fast-diffusing” species) peak is likely to involve Li in a more symmetric environment. It is similar in position and shape (although slightly broader) than that of Li in LiPF₆ and other LiX salts and it may be assigned to solvated Li⁺. The solvated Li⁺ can be “free”, Li⁺(G₄)_n, but not necessarily in a 1/1 molar ratio with G₄, in the form of a contact ion pair (–NSO₂CF₃⁻⋯Li⁺⋯G₄), or even a “free” Li that moves with the rest of the cage (Fig. 7). The broader (“slow-diffusing” species) peak is then assigned to Li that has not disassociated from the POSS cage. Their relative populations based on the integrated intensities of the static ⁷Li NMR peaks is 33/67 (fast/slow), whereas the ratio in a two component fit of the decay in the ⁷Li NMR PFG-NMR experiment is 48/52 (fast/slow). The two ratios are different because the broad peak (“slow-diffusing” species) has a significantly faster relaxation.

Fig. 7 Schematic of G₄/POSS-(LiNSO₂CF₃)₈ showing contact ion pairs, solvent separated ion pairs and un-dissociated Li.

Further insight can be gained by analyzing the radii of the solvated ions, by: (i) the Stokes–Einstein equation originally defined for hard spheres:

where η is the solvent viscosity and R_s is the effective hydrodynamic (Stokes) radius, a relation that was found to hold for LiX in a series of 12 solvents;^48,49 or (ii) analyzing the data by defining an experimental R_ion, which corresponds to the Stokes radius of the ions with respect to the solvent:

Radii obtained by both methods are presented in Table 1, using measured solution viscosities (Table S3, ESI†) for the Stoke–Einstein equation.

The transference numbers were calculated from self-diffusion coefficients (Table 1) using eqn (5) and (6):

where D_Li and D_PF6 or D_8mer (8mer = POSS-(LiNSO₂CF₃)₈) are the diffusion coefficients of any species containing Li or F nuclei, respectively, and χ_PF6 and χ_8mer are the mole fractions of LiPF₆ and POSS-(LiNSO₂CF₃)₈, respectively, in the mixture. Comparison of the lithium ion transference numbers obtained by potentiostatic titration and PFG-NMR indicates that they are not the same (Table 1). In all cases we observe that t^PP₊ > t^NMR₊, although what is commonly observed is that t^NMR₊ > t^PP₊, since measurement by PFG-NMR includes charged and neutral species.

In order to better understand the measured diffusion coefficients, associated radii and transference numbers, the actual species present in solution must be considered. In particular, we want to elucidate why t^PP₊ > t^NMR₊ and why there is only one ¹⁹F NMR signal and two ⁷Li NMR signals. In general, lithium ion transference numbers depend on the possible species and their relative amounts in solution. For simple LiX salts in aprotic polar solvents, LiX ↔ Li⁺ + X⁻ equilibria exist, so that free (solvated) ions, contact ion pairs (CIPs) and higher order multiplets can be present. In most treatments, only free solvated ions and CIPs are considered.

For LiPF₆, the smaller value of R_s (eqn (3)) for PF₆⁻ than for Li⁺ indicates that there is less interaction between the anion and G₄ than between Li⁺ and G₄. That is, G₄ preferentially solvates Li⁺ ions compared with the PF₆⁻ anions, as also reported for G₄/Li triflate (Tf = LiCF₃SO₃) [R_Li ∼ 0.317–0.383 nm and R_Tf ∼ 0.289–0.326 nm]. The same trends for LiPF₆ are observed using eqn (4), where the ionic radii are normalized against the solvent radius.

In the case of POSS-(LiNSO₂CF₃)₈, there are many possible multiply charged as well as neutral species.

For POSS-(LiNSO₂CF₃)₈, abbreviated here as PLi₈:

PLi₈ + nXG₄ ↔ x₁PLi₇¹⁻ + x₂PLi₆²⁻ + x₃PLi₅³⁻ + x₄PLi₄⁴⁻ + x₅PLi₃⁵⁻ + x₆PLi₂⁶⁻ + x₇PLi₁⁷⁻ + x₈PLi₀⁸⁻ + XLi⁺(G₄)_n (7)

where x_i is the mole fraction of PLi₈₋₁ⁱ⁻, and n is the number of solvent molecules.

If there is rapid exchange between all the Li containing species, the diffusion coefficient measured by PFG-NMR is:

where .

In the case of complete dissociation,

D^NMR_Li = D(Li⁺(G₄)_n) (9)

The corresponding conductivities are:

σ = [ce²/k_BT][8D(Li⁺(G₄)_n) + 64D(PLi₀⁸⁻)] for complete dissociation

where c = salt concentration; e = electric charge, k_B = Boltzmann constant and T is absolute temperature.

The Li⁺ ion transference numbers are then:

The value of the radius obtained from the ¹⁹F PFG-NMR diffusion coefficient (eqn (3)) indicates that R_POSS is small (0.63 nm), in the range of sizes found for POSS compounds (where radii of ∼0.5–0.6 nm are expected from the geometry of the cubes). Since a trimer composed of POSS-(LiNSO₂CF₃)₈⋯Li⋯POSS-(LiNSO₂CF₃)₈ would have a size much larger than those measured in the ¹⁹F PFG-NMR experiment, aggregates of this type are rare and/or not long lived.

It is not possible to obtain lithium ion transference >0.5 using eqn (10) or (11) with any combination of PLi_8−iⁱ⁻ and free Li⁺(G₄)_n. Eqn (11) indicates that for fully dissociated PLi₈, i.e. PLi₀⁸⁻ + 8Li⁺(G₄)_n, the diffusion coefficient of Li⁺(G₄)_n must be very much greater than 8D(PLi₀⁸⁻) to achieve a high Li⁺ transference number, and there would be only one peak in the ⁷Li NMR spectrum. The radius of this solvated Li⁺ ion would also be expected to be approximately the same size as that observed in G₄/LiPF₆, ∼0.33 nm, and in G₄/Li triflate (Tf = LiCF₃SO₃) ∼ 0.317–0.383 nm,⁴⁹ rather than 0.92–1.34 nm measured. Since these effects are not observed, the PLi₈ must be only partially dissociated. Even in this case, i.e. for a distribution of PLi_8−iⁱ⁻, since the two Li peaks have similar areas, this interpretation indicates that the average charge 〈i〉 of the PLi_8−iⁱ⁻ must be significantly greater than unity, assuming that the PLi_8−iⁱ⁻ all have the same average diffusion coefficient, 〈D_P〉. The diffusion coefficient of Li⁺(G₄)_n, D(Li⁺(G₄)_n), must then be greater than about 5 times 〈D_P〉 to achieve a high Li⁺ ion transference number. However, the diffusion coefficients obtained from ⁷Li PFG-NMR are smaller for the two Li pools than the diffusion coefficient obtained by ¹⁹F PFG-NMR for the PLi_8−iⁱ⁻, so that the Li⁺ ion transference number would be substantially less than 0.5.

In order to obtain lithium ion transference numbers >0.5, the peaks in the Li PFG-NMR must be assigned to solvated and un-solvated Li⁺ ions. The peak at 1.34 nm is assigned to PLi_8−iⁱ⁻ in which the Li is un-dissociated. The solvated Li⁺ peak (0.92 nm) is composed of free Li⁺(G₄)_n and PLi_8−iⁱ⁻·Li(G₄)_n, i.e. PLi_8−iⁱ⁻ solvated with Li⁺. The PLi_8−iⁱ⁻·Li(G₄)_n can be contact ion pairs or Li⁺(G₄)_n moving with the anion; the latter can arise if the anion and cation fluxes are coupled (Fig. 7). Since PFG-NMR measures the average diffusion coefficient of the nuclei in each pool, the diffusion coefficient of the PLi_8−iⁱ⁻·Li⁺(G₄)_n (which do not contribute to Li⁺ ion conductivity since they are charge neutral) may be substantially smaller than the diffusion coefficient of the “free” solvated Li⁺ ions (which are the only lithium ions that contribute to the Li⁺ ion conductivity) and result in the (average) small diffusion coefficients measured.

In addition, the solvated lithium ions that are linked to the PLi_8−iⁱ⁻ ions reduce their effective charges, and hence their contribution to the anion conductivity. In other words, the dissolved PLi₈ exists as {PLi_8−iⁱ⁻·(LiG₄)_j}^(i−j)−, and the effective degree of dissociation is small, thus reducing the anion conductivity. This makes PLi₈ act very much like a 1 : 1 salt in solution from a conductivity point of view. A small effective degree of dissociation enables the diffusion coefficient of the free solvated Li⁺ ions to be substantially larger than the average diffusion coefficient of all the solvated Li⁺ ions (that also include the slower moving Li⁺ associated with the large anion). If the diffusion coefficient of the free solvated Li⁺ ions is substantially larger than the average diffusion coefficient of all the solvated Li⁺ ions, a Li⁺ ion transference number greater than 0.5 can be achieved. A small degree of effective dissociation can also account for the much lower conductivity of POSSLi₈ relative to simple lithium salts despite the high PFG NMR diffusion coefficients.

In the ¹⁹F NMR, there is broadening of the CF₃ resonance compared with the narrower PF₆⁻ of the LiX salt, indicating a wider distribution of anion populations or environments that could arise from a range of solvated and non-solvated Li⁺ ions associated with the anion (but still with only a single ¹⁹F NMR peak). As in the case of other LiX salts, when contact ion pairs exist, the effect of un-dissociated dimers on the average diffusion constant (〈X⁻ + LiX〉) of the anion (i.e. addition to a small Li to a larger X⁻) is smaller than the effect of un-dissociated dimers on the average diffusion constant (〈Li⁺ + LiX〉) of Li (i.e. addition of large X⁻ to small Li)].⁴⁹ This explains why there can be one ¹⁹F NMR signal but two ⁷Li NMR signals.

The association of Li⁺ ions with G₄ and the anion are dynamic processes that are somewhat reminiscent of the Grotthuss mechanism in water. However, here charge does not move from one Li atom to another, so it is better described as coordinated hopping of Li⁺. The Li⁺ ion is associated dynamically with an average of n solvent molecules and/or anions. Thus D(Li⁺(G₄)_n) in the above equations should be interpreted as the diffusion coefficient of Li⁺ ions that are instantaneously but not permanently solvated by an average of n solvent molecules, and where one solvent molecule can displace another. In a contact ion pair, the –NSO₂CF₃⁻⋯Li⁺⋯G₄ contact can be replaced with an oxygen from G₄(G₄⋯Li⁺) forming Li⁺(G₄)_n. The diffusion of Li⁺ ions can then occur either by a Grotthuss type (coordinated hopping) mechanism (rapid exchange between species) or translation (diffusion) of a complex of Li⁺(G₄)_n (if species long-lived). Grotthuss type mechanisms of Li⁺ ion transfer, where charge transport has been decoupled from mass transport, have been modelled using molecular dynamics simulations (MDS) of concentrated solutions of acrylonitrile (AN)–LiX, consisting of dissociated ions, neutral contact ion pairs (CIPs) or neutral aggregates (AGG).⁵⁰ The coordinated hopping of Li⁺ between ion pairs or higher order clusters has been predicted by MDS,⁵¹ where hopping between aggregates is unlikely unless they are closely spaced.⁵² Therefore, this mechanism can occur between two POSS-(LiNSO₂CF₃)₈ cubes only if they are close enough together. For G₄/POSS-(LiNSO₂CF₃)₈ O/Li = 20/1, the center-to-center distance is estimated as 2.27 nm (cubic array)/2.41 nm (FCC) so that the largest size (1.13 nm radius) measured for the Li species by ⁷Li PFG-NMR is the closest possible approach for two contacting spheres. The SAXS data for the more concentrated G₄/POSS-(LiNSO₂CF₃)₈ O/Li = 7.5 sample indicates that the spheres can be overlapping.

In the mixed system, the radii of Li⁺ ions are almost the same in G₄/LiPF₆ and mixed G₄/LiPF₆/POSS-(LiNSO₂CF₃)₈ samples, and the radii of the PF₆⁻ ions are also almost the same in G₄/LiPF₆ and mixed G₄/LiPF₆/POSS-(LiNSO₂CF₃)₈ samples, suggesting that each ion is similarly solvated in both systems. Since the POSS-(LiNSO₂CF₃)₈ is expected to become more dissociated with dilution, more dissociated Li⁺ ions solvated by a single G₄ are expected. Further, in mole ratios, the [LiPF₆] = 85 [POSS-(LiNSO₂CF₃)₈], so the dominant species is LiPF₆. However, the radius of the POSS-(LiNSO₂CF₃)₈ anion measured by ¹⁹F PFG-NMR increases, but not to twice the size expected if triplets were formed. We thus suggest that the POSS anions form complexes with the PF₆⁻ anion, i.e. –NSO₂CF₃⁻⋯Li⁺⋯PF₆⁻. This effectively slows its diffusion constant and permits a higher transference number measured by potentiostatic polarization.

Electrochemical characterization

Electrochemical stability and reversibility

The anodic stability of G₄/POSS-(LiNSO₂CF₃)₈, G₄/(LiTFSI + 20% POSS-(LiNSO₂CF₃)₈), and G₄/LiTFSI was measured with linear sweep voltammetry at 25 °C from open circuit voltage (OCV) to 5 V at a scan rate of 2 mV s⁻¹, using stainless steel working electrodes and lithium counter/reference electrodes (Fig. S7, ESI†). The stability window was about the same (2 V to 4.1–4.3 V) for the three electrolytes, using a threshold current of 50 μA cm⁻² typical for glyme electrolytes,³⁸ and thus the electrolytes were stable in the voltage range (2.6 to 4 V) used for half-cell testing with the LiFePO₄ cathode. The time dependence of the resistance under open circuit voltage at 25 °C (Fig. S8, ESI†) was obtained using a Li⁰/electrolyte/Li⁰ cell for G₄/POSS-(LiNSO₂CF₃)₈ and G₄/(80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈) with electrochemical impedance spectroscopy. The bulk resistance was constant for both electrolytes, with the bulk resistance of G₄/POSS-(LiNSO₂CF₃)₈ greater than that of G₄/[80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈], as expected from the conductivity data. There was a gradual but slight decrease in the interfacial resistance for [G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈], but for G₄/POSS-(LiNSO₂CF₃)₈ the interfacial resistance dropped steeply for several days, which may be due to improved wetting of the electrodes with time or to decreased diffusion in the more viscous G₄/POSS-(LiNSO₂CF₃)₈, and then leveled off to the same value as for the mixed system. This suggests that POSS-(LiNSO₂CF₃)₈ affects the formation of the interfacial layer on Li⁰ metal and its resistance similarly for both systems, and may be due to the formation of a particle rich coating on the Li⁰ metal, as previously suggested for hairy nanoparticles.⁵³

Lithium plating and stripping

Lithium plating and stripping for Li⁰/[G₄/LiTFSI]/Li⁰, Li⁰/[G₄/POSS-(LiNSO₂CF₃)₈/Li⁰ and Li⁰/[G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈]/Li⁰ were studied at different current densities (0.01 to 1 mA cm⁻²) for 2 hour charge and discharge at 25 °C. Cell failure occurred in the order G₄/80% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈ < G₄/LiTFSI < G₄/POSS-(LiNSO₂CF₃)₈, (although not all cells were cycled until failure). At 0.01 mA cm⁻², G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈ and G₄/LiTFSI exhibited stable cycling, but for G₄/POSS-(LiNSO₂CF₃)₈, the polarization increased with cycle number (Fig. S9, ESI†). At 0.1 mA cm⁻², both G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈ and G₄/LiTFSI exhibited cyclic stability and stable voltage profiles for 11 days (when cycling was stopped) (Fig. 8A), and the overvoltage was less in the case of the G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈. At 1 mA cm⁻² (Fig. S9, ESI†), cell failure occurred at <1 cycle, ∼1 day and was stopped at 5 days, for G₄/POSS-(LiNSO₂CF₃)₈, G₄/LiTFSI and G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈, respectively, although polarization increased with cycle number for G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈.

Fig. 8 Electrochemical results (all cells O/Li = 20/1): (A) comparison of cycling data using 0.1 mA cm⁻² and 2 h charge/2 h discharge for Li⁰/[G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈]/Li⁰ and Li⁰/(G₄/LiTFSI)/Li⁰; (B) rate capability Li⁰/[G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈]/LiFePO₄ at C/20, C/10, C/5, C/2, 1C, 2C, C/10 and C/20; (C) discharge capacity and Coulombic efficiency as a function of cycle number for Li⁰/[G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈]/LiFePO₄ cell at C/4 rate and 25 °C; inset shows cell voltage vs. specific capacity at 1, 10, 50 and 100 cycles; (D) comparison of discharge capacity for Li⁰/[G₄/POSS-(LiNSO₂CF₃)₈]/LiFePO₄ and Li⁰/(G₄/LiTFSI)/LiFePO₄ at C/5 rate.

Half-cell testing

Cyclic voltammetry (CV) was used to study the reversibility of the Li/(G₄/POSS-(LiNSO₂CF₃)₈ O/Li = 20/1)/LiFePO₄ system. (Fig. S10, ESI†); the preparation of the LiFePO₄ electrode is provided in ESI.† For practical applications, the battery should be able to run at different C-rates with high capacity. The galvanostatic performance of half cells using Li⁰ metal as the anode and LiFePO₄ as the cathode were investigated for the G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈, G₄/LiTFSI and G₄/POSS-(LiNSO₂CF₃)₈ electrolytes as a function of C-rate. The high ionic conductivity of electrolytes (0.25 mS cm⁻¹ to 3 mS cm⁻¹) allows the batteries to be tested at RT. The expected decrease in capacity with C-rate for Li⁰/[G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈]/LiFePO₄ (Fig. 8B), which recovers when returned to low C-rates, demonstrates the stability of the electrolyte throughout the testing. Stable performance was demonstrated at C/4, with retention of 130 mA h g⁻¹ until the 100th cycle, when cycling was stopped (Fig. 8C). Better cycling behavior was demonstrated for the G₄/POSS-(LiNSO₂CF₃)₈ than for the G₄/LiTFSI electrolyte at C/5 (Fig. 8D): the former exhibited stable cycling until the 100th cycle, while the latter showed capacity-fade after the 50th cycle. At 3C, capacity fade began at ∼30 cycles (Fig. S11, ESI†), and decreased slightly less for the G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈ than for the G₄/LiTFSI electrolyte. By comparison, for 1 M LiPF₆ in a mixture of ethylene carbonate (EC) and ethyl methyl carbonate (EMC) (EC : EMC 4 : 6 by weight) and a Li⁰/NCA cell (NCA = LiNi_0.80Co_0.15Al_0.05O₂), the specific capacity started decreasing after 15 cycles at 1C, 2C and 3C rates.⁵⁴

The cause of cell failure or capacity fade was not directly ascertained. However, it is interesting to note that recent work on dendritic growth in liquid electrolytes has shown that this growth is inhibited by electrolytes with low viscosities and large anions (with little effect of conductivity).⁵⁵ In the current investigation, cells lasted longest and had the best half-cell cycling using the electrolyte that had large anions as well as low viscosity (G₄/80 wt% LiTFSI/20 wt% POSS-(LiNSO₂CF₃)₈). The role of nanoparticles in stabilizing the interface and reducing interfacial resistance of Li⁰ metal has also been demonstrated.⁵³

Conclusions

In summary, we have shown that multi-ionic lithium salts (POSS-(LiNSO₂CF₃)₈, comprised of rigid polyoligomeric silsesquioxanes (POSS) cubes functionalized with eight lithium[(4-styrenesulfonyl)trifluoromethane-sulfonyl)imide] groups (LiNSO₂CF₃), can be added to solutions of LiTFSI or LiPF₆ in tetraglyme (G₄), increasing the lithium ion transference number but maintaining high ionic conductivity. The (POSS-(LiNSO₂CF₃)₈ are only partially dissociated, and better ionic conductivity is expected with more dissociative pendant anions. Unlike single ion conducting (SIC) polymers added to electrolyte solutions, addition of the compact POSS-(LiNSO₂CF₃)₈ provides a source of less mobile anions without the same large increase in viscosity as in the flexible SIC polymers, and avoids the formation of clusters/aggregates that trap Li⁺ ions and prevent high conductivity in these SIC systems. In neat G₄/POSS-(LiNSO₂CF₃)₈, solvent-in-salt electrolytes are formed at high POSS-(LiNSO₂CF₃)₈ concentration and stable colloids are formed with increasing solvent. When the POSS-(LiNSO₂CF₃)₈ are in close proximity to each other in G₄ solutions, there are many possible rapidly equilibrating Li species such as contact ion pairs and solvent separated ion pairs that can result in the transport of Li⁺ ions by Grotthuss type coordinated hopping mechanisms as well as by diffusion.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

Financial support of this work from NSF CBET award 1437814 is gratefully acknowledged. We thank Dr Gerd Langenbucher of Anton-Paar, Inc. for help in obtaining the SAXS data.

References

1. M. Armand and J. M. Tarascon, Building better batteries, Nature, 2008, 451(7179), 652–657.
2. P. G. Bruce, S. A. Freunberger, L. J. Hardwick and J. M. Tarascon, Li-O(2) and Li-S batteries with high energy storage, Nat. Mater., 2012, 11(1), 19–29.
3. L. Grande, E. Paillard, J. Hassoun, J. B. Park, Y. J. Lee, Y. K. Sun, S. Passerini and B. Scrosati, The Lithium/Air Battery: Still an Emerging System or a Practical Reality?, Adv. Mater., 2015, 27(5), 784–800.
4. C. Monroe and J. Newman, The impact of elastic deformation on deposition kinetics at lithium/polymer interfaces, J. Electrochem. Soc., 2005, 152(2), A396–A404.
5. J. N. Chazalviel, Electrochemical aspects of the generation of ramified metallic electrodeposits, Phys. Rev. A: At., Mol., Opt. Phys., 1990, 42(12), 7355–7367.
6. M. Doyle, T. F. Fuller and J. Newman, The importance of the lithium ion transference number in lithium/polymer cells, Electrochim. Acta, 1994, 39(13), 2073–2081.
7. K. M. Diederichsen, E. J. McShane and B. D. McCloskey, Promising Routes to a High Li+ Transference Number Electrolyte for Lithium Ion Batteries, ACS Energy Lett., 2017, 2563–2575.
8. P. G. Bruce, J. Evans and C. A. Vincent, Conductivity and transference number measurements on polymer electrolytes, Solid State Ionics, 1988, 28, 918–922.
9. N. P. Balsara and J. Newman, Relationship between Steady-State Current in Symmetric Cells and Transference Number of Electrolytes Comprising Univalent and Multivalent Ions, J. Electrochem. Soc., 2015, 162(14), A2720–A2722.
10. S. Zugmann, M. Fleischmann, M. Amereller, R. M. Gschwind, H. D. Wiemhöfer and H. J. Gores, Measurement of transference numbers for lithium ion electrolytes via four different methods, a comparative study, Electrochim. Acta, 2011, 56(11), 3926–3933.
11. P. G. Bruce and C. A. Vincent, Steady state current flow in solid binary electrolyte cells, J. Electroanal. Chem., 1987, 225(1–2), 1–17.
12. J. Evans, C. A. Vincent and P. G. Bruce, Electrochemical measurement of transference numbers in polymer electrolytes, Polymer, 1987, 28(13), 2324–2328.
13. E. L. Hahn, Spin echos, Phys. Rev., 1950, 80(4), 580–594.
14. E. O. Stejskal and J. E. Tanner, Spin diffusion measurements: spin echos in the presence of a time-dependent field gradient, J. Chem. Phys., 1965, 42(1), 288–292.
15. P. T. Callaghan, Principles of Nuclear Magnetic Resonance, Oxford University Press, Oxford, 1991.
16. S. Abbrent and S. Greenbaum, Recent progress in NMR spectroscopy of polymer electrolytes for lithium batteries, Curr. Opin. Colloid Interface Sci., 2013, 18(3), 228–244.
17. H. Zhang, C. M. Li, M. Piszcz, E. Coya, T. Rojo, L. M. Rodriguez-Martinez, M. Armand and Z. B. Zhou, Single lithium-ion conducting solid polymer electrolytes: advances and perspectives, Chem. Soc. Rev., 2017, 46(3), 797–815.
18. S. Mogurampelly, O. Borodin and V. Ganesan, Computer Simulations of Ion Transport in Polymer Electrolyte Membranes, in Annual Review of Chemical and Biomolecular Engineering, ed. J. M. Prausnitz, Annual Reviews, Palo Alto, 2016, vol. 7, pp. 349–371.
19. M. F. Lu, J. Runt and P. Painter, An Infrared Spectrocopic Study of a Polyester Copolymer Ionomer Based on Poly(ethylene oxide), Macromolecules, 2009, 42(17), 6581–6587.
20. Y. V. Oza, D. R. MacFarlane, M. Forsyth and L. A. O'Dell, Unexpected effect of tetraglyme plasticizer on lithium ion dynamics in PAMPS based ionomers, Phys. Chem. Chem. Phys., 2016, 18(28), 19011–19019.
21. Y. Y. Lu, M. Tikekar, R. Mohanty, K. Hendrickson, L. Ma and L. A. Archer, Stable Cycling of Lithium Metal Batteries Using High Transference Number Electrolytes, Adv. Energy Mater., 2015, 5(9), 7.
22. H. Oh, K. Xu, H. D. Yoo, D. S. Kim, C. Chanthad, G. Yang, J. Z. Jin, I. A. Ayhan, S. M. Oh and Q. Wang, Poly(arylene ether)-Based Single-Ion Conductors for Lithium-Ion Batteries, Chem. Mater., 2016, 28(1), 188–196.
23. L. Porcarelli, A. S. Shaplov, F. Bella, J. R. Nair, D. Mecerreyes and C. Gerbaldi, Single-Ion Conducting Polymer Electrolytes for Lithium Metal Polymer Batteries that Operate at Ambient Temperature, ACS Energy Lett., 2016, 1(4), 678–682.
24. R. Meziane, J.-P. Bonnet, M. Courty, K. Djellab and M. Armand, Single-ion polymer electrolytes based on a delocalized polyanion for lithium batteries, Electrochim. Acta, 2011, 57, 14–19.
25. S. W. Feng, D. Y. Shi, F. Liu, L. P. Zheng, J. Nie, W. F. Feng, X. J. Huang, M. Armand and Z. B. Zhou, Single lithium-ion conducting polymer electrolytes based on poly(4-styrenesulfonyl)(trifluoromethanesulfonyl)imide anions, Electrochim. Acta, 2013, 93, 254–263.
26. Q. Ma, H. Zhang, C. W. Zhou, L. P. Zheng, P. F. Cheng, J. Nie, W. F. Feng, Y. S. Hu, H. Li, X. J. Huang, L. Q. Chen, M. Armand and Z. B. Zhou, Single Lithium-Ion Conducting Polymer Electrolytes Based on a Super-Delocalized Polyanion, Angew. Chem., Int. Ed., 2016, 55(7), 2521–2525.
27. J. L. Schaefer, D. A. Yanga and L. A. Archer, High Lithium Transference Number Electrolytes via Creation of 3-Dimensional, Charged, Nanoporous Networks from Dense Functionalized Nanoparticle Composites, Chem. Mater., 2013, 25(6), 834–839.
28. K. D. Kreuer, A. Wohlfarth, C. C. de Araujo, A. Fuchs and J. Maier, Single Alkaline-Ion (Li+, Na+) Conductors by Ion Exchange of Proton-Conducting Ionomers and Polyelectrolytes, ChemPhysChem, 2011, 12(14), 2558–2560.
29. M. Videa, W. Xu, B. Geil, R. Marzke and C. A. Angell, High Li+ self-diffusivity and transport number in novel electrolyte solutions, J. Electrochem. Soc., 2001, 148(12), A1352–A1356.
30. H. G. Buss, S. Y. Chan, N. A. Lynd and B. D. McCloskey, Nonaqueous Polyelectrolyte Solutions as Liquid Electrolytes with High Lithium Ion Transference Number and Conductivity, ACS Energy Lett., 2017, 2(2), 481–487.
31. L. Suo, Y.-S. Hu, H. Li, M. Armand and L. Chen, A new class of Solvent-in-Salt electrolyte for high-energy rechargeable metallic lithium batteries, Nat. Commun., 2013, 4, 1481.
32. P. Johansson, J. Tegenfeldt and J. Lindgren, Modelling amorphous lithium salt-PEO polymer electrolytes: ab initio calculations of lithium ion-tetra-, penta- and hexaglyme complexes, Polymer, 1999, 40(15), 4399–4406.
33. S. Tsuzuki, W. Shinoda, M. Matsugami, Y. Umebayashi, K. Ueno, T. Mandai, S. Seki, K. Dokko and M. Watanabe, Structures of Li(glyme) (+) complexes and their interactions with anions in equimolar mixtures of glymes and Li TFSA: analysis by molecular dynamics simulations, Phys. Chem. Chem. Phys., 2015, 17(1), 126–129.
34. G. M. Mao, M. L. Saboungi, D. L. Price, M. B. Armand and W. S. Howells, Structure of liquid PEO-LiTFSI electrolyte, Phys. Rev. Lett., 2000, 84(24), 5536–5539.
35. T. Tamura, K. Yoshida, T. Hachida, M. Tsuchiya, M. Nakamura, Y. Kazue, N. Tachikawa, K. Dokko and M. Watanabe, Physicochemical Properties of Glyme-Li Salt Complexes as a New Family of Room-temperature Ionic Liquids, Chem. Lett., 2010, 39(7), 753–755.
36. K. Yoshida, M. Tsuchiya, N. Tachikawa, K. Dokko and M. Watanabe, Change from Glyme Solutions to Quasi-ionic Liquids for Binary Mixtures Consisting of Lithium Bis(trifluoromethanesulfonyl)amide and Glymes, J. Phys. Chem. C, 2011, 115(37), 18384–18394.
37. C. Zhang, A. Yamazaki, J. Murai, J. W. Park, T. Mandai, K. Ueno, K. Dokko and M. Watanabe, Chelate Effects in Glyme/Lithium Bis(trifluoromethanesulfonyl)amide Solvate Ionic Liquids, Part 2: Importance of Solvate-Structure Stability for Electrolytes of Lithium Batteries, J. Phys. Chem. C, 2014, 118(31), 17362–17373.
38. T. M. Pappenfus, W. A. Henderson, B. B. Owens, K. R. Mann and W. H. Smyrl, Complexes of lithium imide salts with tetraglyme and their polyelectrolyte composite materials, J. Electrochem. Soc., 2004, 151(2), A209–A215.
39. K. Hayamizu, E. Akiba, T. Bando and Y. Aihara, ¹H, ⁷Li and ¹⁹F nuclear magnetic resonance and ionic conductivity studies for liquid electrolytes composed of glymes and polyethyleneglycol dimethyl ethers of CH₃O(CH₂CH₂O)_nCH₃ (n = 3–50) doped with LiN(SO₂CF₃)₂, J. Phys. Chem., 2002, 117(12), 5929–5939.
40. L. Aguilera, S. Z. Xiong, J. Scheers and A. Matic, A structural study of LiTFSI-tetraglyme mixtures: From diluted solutions to solvated ionic liquids, J. Mol. Liq., 2015, 210, 238–242.
41. K. Shimizu, A. A. Freitas, R. Atkin, G. G. Warr, P. A. FitzGerald, H. Doi, S. Saito, K. Ueno, Y. Umebayashi, M. Watanabe and J. N. C. Lopes, Structural and aggregate analyses of (Li salt plus glyme) mixtures: the complex nature of solvate ionic liquids, Phys. Chem. Chem. Phys., 2015, 17(34), 22321–22335.
42. K. Shimizu, C. E. S. Bernardes, A. Triolo and J. N. C. Lopes, Nano-segregation in ionic liquids: scorpions and vanishing chains, Phys. Chem. Chem. Phys., 2013, 15(38), 16256–16262.
43. J. L. Nowinski, P. Lightfoot and P. G. Bruce, Structure of LiN(CF₃SO₂)₂, a Novel Salt for Electrochemistry, J. Mater. Chem., 1994, 4(10), 1579–1580.
44. K. Larsson, The crystal structure of octa-(silsesquioxane)(HSiO_1.5)₈, Ark. Kemi, 1960, 16(3), 215–219.
45. F. Wohde, M. Balabajew and B. Roling, Li+ Transference Numbers in Liquid Electrolytes Obtained by Very-Low-Frequency Impedance Spectroscopy at Variable Electrode Distances, J. Electrochem. Soc., 2016, 163(5), A714–A721.
46. P. G. Bruce, M. T. Hardgrave and C. A. Vincent, Steady-state current flow in solid binary electrolyte cells. 2. The effect of ion association, J. Electroanal. Chem., 1989, 271(1–2), 27–34.
47. P. G. Bruce and C. A. Vincent, Effect of ion association on transport in polymer electrolytes, Faraday Discuss., 1989, 88, 43–54.
48. K. Hayamizu, Y. Aihara, S. Arai and W. S. Price, Self-diffusion coefficients of lithium, anion, polymer, and solvent in polymer gel electrolytes measured using Li-7, F-19, and H-1 pulsed-gradient spin-echo NMR, Electrochim. Acta, 2000, 45(8–9), 1313–1319.
49. K. Hayamizu, Y. Aihara, S. Arai and C. G. Martinez, Pulse-gradient spin-echo H-1, Li-7, and F-19 NMR diffusion and ionic conductivity measurements of 14 organic electrolytes containing LiN(SO₂CF₃)₂, J. Phys. Chem. B, 1999, 103(3), 519–524.
50. D. M. Seo, O. Borodin, D. Balogh, M. O'Connell, Q. Ly, S. D. Han, S. Passerini and W. A. Henderson, Electrolyte Solvation and Ionic Association III. Acetonitrile-Lithium Salt Mixtures-Transport Properties, J. Electrochem. Soc., 2013, 160(8), A1061–A1070.
51. K. R. Lu, J. K. Maranas and S. T. Milner, Ion-mediated charge transport in ionomeric electrolytes, Soft Matter, 2016, 12(17), 3943–3954.
52. M. Doyle, M. E. Lewittes, M. G. Roelofs and S. A. Perusich, Ionic conductivity of nonaqueous solvent–swollen ionomer membranes based on fluorosulfonate, fluorocarboxylate, and sulfonate fixed ion groups, J. Phys. Chem. B, 2001, 105(39), 9387–9394.
53. S. Choudhury, A. Agrawal, S. Y. Wei, E. Jeng and L. A. Archer, Hybrid Hairy Nanoparticle Electrolytes Stabilizing Lithium Metal Batteries, Chem. Mater., 2016, 28(7), 2147–2157.
54. D. P. Lv, Y. Y. Shao, T. Lozano, W. D. Bennett, G. L. Graff, B. Polzin, J. G. Zhang, M. H. Engelhard, N. T. Saenz, W. A. Henderson, P. Bhattacharya, J. Liu and J. Xiao, Failure Mechanism for Fast-Charged Lithium Metal Batteries with Liquid Electrolytes, Adv. Energy Mater., 2015, 5(3), 7.
55. M. S. Park, S. B. Ma, D. J. Lee, D. Im, S. G. Doo and O. Yamamoto, A Highly Reversible Lithium Metal Anode, Sci. Rep., 2014, 4, 8.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c7mh01130j

---