Alternative solid-electrolytes are the next key step in advancing lithium batteries with better thermal and chemical stability. A soft-solid electrolyte (Adpn)2LiPF6 (Adpn = adiponitrile) is synthesized and characterized, which exhibits high thermal and electrochemical stability and good ionic conductivity, overcoming several limitations of conventional organic and ceramic materials. The surface of the electrolyte possesses a liquid nano-layer of Adpn that links grains for a facile ionic conduction without high pressure/temperature treatments. Further, the material can quickly self-heal if fractured and provides liquid-like conduction paths via the grain boundaries. A significantly high ion conductivity (~ 10 -4 S/cm) and lithium-ion transference number (0.54) are obtained due to weak interactions between "hard" (charge-dense) Li + ions and "soft" (electronically polarizable) -C≡N group of Adpn. Molecular simulations predict that Li + ions migrate at the co-crystal grain boundaries with a (preferentially) lower Ea and within the interstitial regions between the co-crystals with higher Ea, where the bulk conductivity comprises a smaller but extant contribution. These cocrystals establish a special concept of crystal design to increase the thermal stability of LiPF6 by separating ions in Adpn solvent matrix, and also exhibit a unique mechanism of ion-conduction via low-resistance grain-boundaries, which is contrasting to ceramics or gel-electrolytes.
Replacement of volatile liquid electrolytes to improve safety in lithium metal or lithium-ion batteries (LIBs) has generated interest in the development of solid electrolytes. These have included polymers, polymer gels with nonvolatile liquids, and a wide range of inorganic materials 1 such as oxide and sulfide-based lithium (LICC) or sodium ion-conducting ceramics. 2 While LICCs can have good ionic conductivities (σ) and lithium-ion transference numbers (tLi + , the fraction of charge carried by Li + ), there are still problems associated with poor interfacial contact between the electrolyte and the electrodes. We have been investigating a new class of solid electrolytes: saltorganic co-crystals (also referred to as solvates) of lithium and sodium salts with weakly ligating molecular organic compounds. Similar to lithium-ion-conducting ceramics, these new "soft" cocrystals also have channels for ion migration but are not necessarily single-ion conductors. Unlike the rigid anionic lattices of ceramic electrolytes, the channels in the "soft" co-crystals consist of weakly Lewis basic donor groups of organic molecules, and in some examples, there are no Li•••anion contacts 3 . Previously reported solvates of lithium salts with polyethylene oxide (PEO) 4,5 or glymes 6 can have ionic conductivities higher than their respective molten phases 5 , but are still very low (10 -7 S/cm for PEO solvates [5][6][7] and 10 -6 S/cm for glyme solvates 8 ) due to the tight chelation of the "hard" (non-polarizable) Li + ions with the "hard" ether oxygen donors. Nitriles and other triple bonded systems are known to be electronically soft due to polarizability, characterizable by the hardness factor. 9 Previous work in other 10 and our groups has focused on co-crystals of lithium (or sodium) salts with a coordinating ligand that has "soft" (polarizable), weakly electron-donating atoms, e.g. dimethyl formamide (DMF) with LiCl 11 or NaClO4 [12][13][14] , isoquinoline with LiCl 15 , or adiponitrile (Adpn) with NaClO4 16 , where the weaker binding promotes higher conductivity. Unlike inorganic ceramics where grain boundaries must be sintered and can be resistive, the grain boundaries in soft-solid cocrystals are fluid, permitting easy reformation upon cracking and can be melted and are thus melt-castable at moderate temperatures. 12,16 In the current work, soft co-crystals formed with LiPF6 salt and Adpn solvent: (Adpn)2LiPF6 are investigated experimentally and theoretically. The components of these co-crystals, LiPF6 and Adpn, have compatibility with high voltage cathodes (Adpn) and are used (LiPF6) in most commercial LIBs despite poor thermal stability in pure or solvated form. Polar nitrile (-C≡N) groups, with high dipole moments and dielectric constants of ~ 30, can solvate lithium ions more weakly than the ether oxygens of polyethylene oxide (PEO) or glymes, and have been incorporated as functional groups in liquids, plasticizers, plastic crystals (particularly succinonitrile (SN) 17,18 ), gels (e.g., polyacrylonitrile (PAN) 19,20 ), PAN polymer-in-salt 21,22 and solid electrolytes used for LIB applications 23 . These materials often have high anodic oxidation potentials (> 4.5 V vs. Li + /Li) and are thus resistant to electrochemical oxidation 24 and therefore have the potential to be used with high voltage cathodes 25 , e.g., Li[Mn, Ni, Co]O2 (NMC). Adpn has also been shown to enable the use of high voltage cathodes when added in small amounts (1%) to other electrolyte solutions, by film formation 26 or strong coordination between the Ni 4+ on Ni-rich cathode surfaces and the nitrile groups 27 . Although nitriles suffer from poor reductive stability 24 they can be used with lower-energy anodes such as Li4Ti5O12 (LTO, 1.55 V vs Li + /Li) 18,28 or with graphitic anodes by addition of SEI forming co-solvents 29 . More recently, it was shown that low concentrations of Adpn (1%) in mixed electrolytes formed stable SEIs on Li 0 metal 27 . Of importance for the current work, the reductive stability of acetonitrile (AN) was improved in concentrated (> 4 M) salt solutions, since all of the acetonitrile molecules were passivated by coordination to Li + ions 30 .
The cocrystalline (Adpn)2LiPF6 has an effective molarity of 4.5 M and it exhibits desirable physical properties, including melt-and press-castability, self-healing, high conductivity for an organic solid electrolyte (~10 -4-S•cm -1 ), and a wide electrochemical stability window (5 V), stable cycling (> 50 cycles at C/20, C/10, C/5 rates) with capacities of 140 mAh-g -1 to 100 mAh-g -1 and > 99% Coulombic efficiencies using Li o /(Adpn)2LiPF6/LiFePO4 half-cells and ~ 96% Coulombic efficiency with LTO/(Adpn)2LiPF6/NMC622. Further, our atomistic simulation models demonstrate an unforeseen mechanism of ion-conduction in these soft-solids where grainboundaries are the most vital contributor to the net ionic conductivity, followed by the Li + defects, which facilitate ion-conduction in the bulk co-crystal.
The parent compound (Adpn)2LiPF6 is prepared by heating commercially available LiPF6 in excess adiponitrile, in which it is sparingly soluble at room temperature 29 . After complete dissolution at 165°C, cooling of the solution yields crystalline (Adpn)2LiPF6 , whose stoichiometry is confirmed by XRD analysis of single crystals (Fig. 1a,b). The experimentally obtained powder XRD patterns of these crystals agree with those obtained theoretically from single crystal data, and for postelectrochemical and melt-recrystallized samples (Fig. 1c,d,e,f). The structure exhibits linear parallel Li + ion channels (Fig. 1b), with a shortest distance of 6.23 Å between two successive Li + ions in the b-crystallographic direction. Each Li + ion is coordinated to four cyano groups, and does not interact with any PF6 -anions. An important consequence of the isolation of the PF6 -anion from the Lewis acidic Li + is the improvement of the thermal stability of the salt; the thermogravimetric analysis (TGA) data (Extended Data Fig. 1 31 ). The behavior of co-crystals in TGA was simulated from MD simulations using model V (vacuum), which predicts that Adpn molecules form a liquid-like layer on the co-crystalline surface at room-temperature (Extended Data Fig. 2a) and evaporate as the cocrystals degrade at high temperature (T > 400 K) (Extended Data Fig. 2b -d). This experimental structure was used to benchmark the periodic model P used in MD simulation, provided in Supplementary Fig. S1 andS2.
Differential scanning calorimetry (DSC) data (Fig. 2a) show that the co-crystals of (Adpn)2LiPF6 reversibly melt at 182 0 C, and recrystalize after slight supercooling to ~150°C. The same crystaline phase is recovered after recrystalization based on PXRD (Figure S1b), demonstrating that the material does not decompose upon melting. For comparison, the SN2LiFSI molecular crystal melts at Tm = 59.5 0 C 10 . In the case of (Adpn)2LiPF6, the presence of a small amount of free (i.e., at least one -C≡N not coordinated to Li + ) Adpn in the cocrystal was observed by DSC (Fig. 2a) and Raman spectroscopy (Fig. 2b). The assignment of "free" and coordinated C≡N peaks was validated using the DFT calculations for a geometry shown in Fig. 2c and theoretical vibrational spectra in Fig. 2d (and in Supplementary Table S1).
A liquid-like surface layer was observed in all SEM images of (Adpn)2LiPF6 both before (in pressed and unpressed polycrystalline crystals (Fig. 2e, Extended Data Fig. 3), and after cycling (Extended Data Fig. 4,5). Liquid Adpn juices out of the crystals on application of pressure in the events like cell manufacturing (Supplementary Fig. S3), which is a precedented behavior. 13 With the application of pressure, individual grains fracture into smaller pieces and fuse to one another.
In the large 100 μm single crystals a mosaic structure of smaller, < 1 μm crystalline domains (Extended Data Fig. 3f) could be observed.
EDX spectra (Fig. 2f,g) support the existence of fluid grain boundaries. Relative X-ray fluorescence peaks for carbon and nitrogen atoms of the Adpn are weaker compared with the fluorine and phosphorus atom peaks of the LiPF6 in the grain boundary region than in the grains, suggesting that under the vacuum of the SEM, Adpn is evaporated, leaving behind LiPF6 salt. Such surface liquid phases are well known and are due to a decrease in lattice energy for molecules near the surface of crystals, and most famously illustrated by the surface liquid water layer in ice. 32 This nanoliquid surface behavior is a general characteristic of this class of electrolytes [11][12][13]16,33 and is supported by the MD simulations performed on the model V (Extended Data Fig. 2).
The structure of the intergranular interface was modeled using model V8g with eight nano-sized grains (1 grain supercell = 5x5x5 unit cells) in a box of 30x30x30 nm 3 (Fig. 3a, details provided in Supplementary Information). The simulations show that within a span of a few nanoseconds, the grains interact at the surface and form a more mobile interfacial layer (Fig. 3b, Supplementary Video 1). The formation of this interfacial layer does not depend on the orientation of the grains or the size of the box, viz. does not require lattice matching. Since "excess" Adpn solvent molecules are trapped/confined between the grains during synthesis which form part of the grainboundaries, a model V2g,sol was constructed, where "excess" Adpn molecules solvate two grains of Adpn2LiPF6. The model V2g,sol shows initial inter-phase contact between the two grains (Fig. 3c), where several Li + and PF6 -ions from the crystal dissolve in the free Adpn solvent molecules (Fig. 3d) to facilitate a better grain boundary conduction (discussed in the mechanism of ion transport subsection) in contrast to ceramic organics, whose grains must be sintered to prevent insulating inter-grain gaps.
The electrochemical impedance spectroscopy (EIS) data (Fig. 4a) for a pressed pellet of (Adpn)2LiPF6 in the temperature range between -10⁰C and 80 ⁰C (below its degradation temperature), show a RT conductivity of σ → 10 -4 S cm -1 (using a constant phase-element resister (CPER)/constant phase-element (CPE) circuit shown in Extended Data Fig. 6a), with an Arrhenius activation energy Ea = 37.2 kJ mol -1 . The variation of the DC current as a function of time (Extended Data Fig. 6b), in a Li(s)/(Adpn)2LiPF6/Li(s) cell, including the correction for the interfacial resistance before polarization (R0) and at steady-state (Rs):
gave tLi + = 0.54. Unlike inorganic ceramics with a stationary anion lattice and mobile Li + sublattice (tLi + → 1), here both the anions and cations are mobile, but with a better tLi + than in typical commercial liquid electrolytes. 34 The oxidative current begins to increase only at 5V (Fig. 4b), confirming the excellent oxidative stability of Adpn 29,35 . While Li metal is thermodynamically unstable with most organic solvents, 30,36 Adpn (with or without LiPF6) remains colorless for weeks to months with an inserted piece of Li 0 metal (Supplementary Fig. S4).
Reversible Li 0 stripping/plating is observed for Li o /(Adpn)2LiPF6/Li o (Fig. 4b). This is an improvement for the 4.5M (Adpn)2LiPF6 over dilute AN/LiTFSI (<3M) where Li stripping is not observed. 30 During repeated lithium stripping, the oxidation peak at ~ 0.5 V (vs Li + /Li) superimposes onto itself, while during lithium plating, the reduction peak at ~ -0.5 V decreases from the 1 st through the 3 rd cycles and then remains stable. This suggests that a stable, ionically conductive SEI layer is being formed during the first three cycles. Interfacial resistance (Fig. 4c) in a Li 0 /(Adpn)2LiPF6/Li 0 cell stabilized at about 900Ω after 3 days. Li plating in the same cell (Fig. 4d) was stable for 20-25 days at low current densities (J = 0.01 mA/cm 2 ) but failed at higher current densities.
Excellent cycling data was obtained for the Li 0 /(Adpn)2LiPF6/LiFePO4 half cells at C/20, C/10, and C/5 rates (~ 140 mAh -1 /g) with little capacity fade (Extended Data Fig. 6e,f) between 2.7 and 4 V. The C/10 data ran for > 70 cycles before capacity fade (Fig. 4e). In previous studies with liquid Adpn/1 M salts, only 20 cycles could be obtained with efficiencies of ~ 97%. 37 Thus, the highly concentrated (Adpn)2LiPF6 improves compatibility with the Li 0 metal. A full cell with LTO/(Adpn)2LiPF6/NMC622 was also investigated to assess the potential for these electrolytes to be used with high voltage cathodes. The C-rate performance of LTO/(Adpn)2LiPF6/NMC622 cells (Fig. 4f) and the cycling performance at C/10 (Extended Data Fig. 6g) between 2V and 4.2V vs Li/Li + shows that the accessible capacity decreased with C-rate, as expected, but recovered when returned to the original C-rates. For cells cycled at C/10, the capacity faded up to 50 % over 100 cycles, but without a catastrophic failure (see Extended Data Fig. 6g). Further, the Coulombic efficiency was steady at ~ 99.6% for the whole 100 cycles. Comparison of the Li 0 / Li 0 , Li 0 /LiFePO4, LTO/NMC622 and NMC622/NMC622 cells suggests that both the anodes and cathodes contribute to capacity fade, with comparable impedances for both anodes and cathodes.
Interfacial resistance (Fig. 4c) in the Li 0 /(Adpn)2LiPF6/Li 0 cell stabilized at about 900Ω after 3 days. Lithium plating in the same cell (Fig. 4d) was stable for 20-25 days at low current densities (J = 0.01 mA/cm 2 ) but failed at higher current densities. Post-mortem analysis showed that the Li surface was black. SEM images (Extended Data Fig. 4) indicate that the SEI was rough (not flat) with a thickness < 5 μm after ~ 30 days, suggesting a compact but mossy SEI layer, with no obvious dendritic growth. EDX analysis (Extended Data Fig. 5) of the surface showed residual (Adpn)2LiPF6 crystals that adhered to the SEI layer, based upon their morphology and dominant signals for C, N, F and P atoms but no signal for O atom. At the SEI layer itself, the major peaks were O (from Li2O) and C (from Adpn), with very little P and F (from PF6 -). While further investigation is required, these peaks suggest the formation of lithium oxides and carbonates. The formation of compounds containing C (such as carbonates), and the scarcity of F, P, agree with predications from electronic structure calculations (Extended Data Fig. 7) that the Adpn reduction could cause solvent degradation leading to SEI layer formation at the Li 0 anode. However, the degradation stops after 3 cycles based on cyclic voltammetry (Fig. 4b), giving a stable, protective SEI.
Since the Adpn2LiPF6 electrolyte has a fluid surface region and contains Adpn solvent between the co-crystalline grains, possibly as a nano-confined liquid, the net ion conduction can have at least these two contributions -at the grain-boundaries (𝜎 𝑔𝑏 ) of the cocrystals and inside the cocrystal (𝜎 𝑏𝑢𝑙𝑘 ) where vacancies form. Hence the net ionic conductivity 𝜎 𝑒𝑓𝑓 can be expressed as:
where
and
In order to address these two important contributions, we calculated the MSD for 1000 Li + ions in the V2g,sol model, where two grains of (Adpn)2LiPF6 are simulated in the presence of excess molecules of Adpn (details provided in Supplementary Information). The MSD vs. time plots for individual Li + ions show two major groups with low and high MSDs and some ions in between, at 300 K (Fig. 5d). The ions falling in these groups belong respectively to the bulk (ordered, less mobile, and thus sub-diffusive) and the surface (solvated, highly mobile, and thus linearly diffusive) part of the crystals. At 325 K, more ions solvate with excess Adpn molecules to become linearly diffusive (Fig. 5e) and less ions remain in the bulk-ordered structure (also visible in Fig. 5b). At 350 K, the individual MSDs which had been spread over a wide range on the y-axis at lower temperature become narrow (Fig. 5f), as all the Li + ions are now surfacesolvated and thus exhibit linear diffusion.
The calculated diffusion coefficients (DLi + ) from V2g,sol and 1.7 M solution models derived from linear regime of MSD plots (Extended Data Fig. 8) are provided in the Table S2, and are fitted to the Arrhenius equation (Fig. 6a). The calculated Ea from V2g,sol model considers all the Li + ions present in both the bulk and surface region, hence it should be comparable to Ea,exp from impedance measurements (37 kJ-mol -1 ). The calculated Ea from 1.7 M solution corresponds to a rough model of grain-boundary solvated Li + ions and considers all the Li + ions present in a nano-confined environment of Adpn, and is thus comparable to Ea,gb. The calculated Ea barrier from the V2g,sol model (comparable to Ea,exp) is 44 kJ-mol -1 which is slightly higher than the experimental Ea,exp (37 kJ-mol -1 ). The Ea,gb is 27 kJ-mol -1 , which suggests that most of the conduction occurs via grain boundaries.
Plane-wave DFT calculations were performed to calculate Ea,bulk using Nudged Elastic Band The calculated Ea barriers from various experimental and theoretical methods (Ea,exp = 37, Ea,MD = 44, Ea,bulk,DFT = 72, Ea,gb,MD = 27, all in kJ-mol -1 ) suggest that the most feasible path for Li + conduction is diffusion via grain boundaries. Ion dynamics from model P, analysis of jump events and other mechanistic paths and corresponding structures are provided in Supplementary Fig. S5 -S11.
There are significant differences in the structures and mechanisms of ion conduction between the traditional solid electrolytes and the soft-solid electrolytes. In contrast to inorganic conductors, the soft-solid crystals of (Adpn)2LiPF6 have an interstitial solvent mediated migration of Li + ions, which we demonstrate is a fluid surface layer on the co-crystals and possibly a supersaturated LiPF6/Adpn solution between the co-crystals, facilitating low resistance ion-conduction at the grain boundaries. Future efforts will explore the incorporation of deliberate defecting strategies with an optimal solvent and anion that increases the number of vacancy or interstitial sites (e.g., by isovalent or aliovalent doping, or introduction of defects by chain ends), to improve the ionic conductivity in these electrolytes.
supervised computational efforts. S. L. W. supervised electrochemistry and characterization efforts. M. J. Z. supervised synthetic and X-ray characterization efforts. P. P., A. V., S. L. W., and The location for migrating Li + ion can be seen in the highlighted spot. A dynamic visualization of the Li + ion migration is presented in Supplementary Video 2.
Calorimeter (DSC) TA Instruments 2920 was used to analyze the melt and crystallization temperatures of the (Adpn)2LiPF6, with the sample in hermetically sealed Tzero aluminum pans.
Samples were scanned from -120 °C to 200 °C at a scan rate of 10 °C.min -1 , under ultra-pure N2 purge. The second cycle of the adiponitrile matrix and (Adpn)2LiPF6 powder was reported out of the two measured cycles. Scanning electron microscope (SEM) data were acquired on a FEI Quanta 450FEG SEM) with energy-dispersive X-ray spectroscopy (EDS) capability (Oxford Aztec Energy Advanced EDS System). Raman spectra were recorded in the 100-3000 cm -1 region at room temperature using a Horiba LabRAM HR Evolution Raman spectrometer, with a resolution of 1.8 cm -1 , an excitation wavelength of 532 nm, 60 mW power, and a grating groove density of 600 gr/mm. Samples were measured with 8 acquisitions, and 2 to 8 seconds each, depending on peak intensity. Microscope photos of the pressed crystals were obtained using a Teslong NTE430 microscope inspection camera.
For the conductivity measurements, polycrystalline (Adpn)2LiPF6 powder incorporated into glass fiber separator (which gave consistent separator thickness and prevented shorting). Alternatively, the separator could be prepared by pressing at 800 psi in a hydraulic crimper in an argon purged glove box. The conductivity measurements were performed in a homemade electrochemical cell placed in a N2 purged, temperature-controlled gas chromatography (GC) oven. Temperaturedependent bulk impedance data was measured by AC electrochemical impedance spectroscopy (EIS) using a Gamry Interface 1000 potentiostat/galvanostat/ZRA in the frequency range 0.1-1MHz in a temperature range between 80 0 C and -10 0 C with increments of 10 0 C. The cell was thermally equilibrated for 30 minutes at each temperature before the bulk impedance was measured during both the cooling and heating cycles.
LiPF6 (151 mg, 1.0 mmols) was dissolved in excess adiponitrile (Adpn) (4.5 ml, 4.0 mmol) by heating the mixture to 165 0 C under an argon atmosphere until it dissolved. The LiPF6 was not soluble in Adpn at RT. Upon cooling, crystalline material started to form at about 115 0 C and was complete at room temperature (RT). A single crystal was removed from the precipitate for X-ray structure determination, and the remaining solid was rinsed five times with excess Et2O and dried under vacuum for ~ 20 minutes to remove the residual amount of Adpn and Et2O, after which the dry, co-crystal solid was isolated; no visible liquid adiponitrile/solvent, was apparent by inspection. Water is rigorously excluded from all samples as we use a glove box with argon atmosphere where water levels are < 9 ppm at worst. The sample was incorporated concomitantly during the synthesis into Whatman glass microfiber filters (GF), grade GF/A (Sigma-Aldrich) 0.26 mm thickness. The co-crystal in the glass fiber filter was used as the separator between the electrodes to control the size and the amount of the electrolyte in the electrochemical experiments.
For all other electrochemical measurements (Li + ion transference numbers, plating and stripping, cyclic voltammetry, linear sweep voltammetry, and full cell cycling) the (Adpn)2LiPF6 was incorporated in the glass filters (during the synthesis). Before use, the lithium metal was polished using Teflon blocks. For the plating and striping experiments at room temperature with the Li 0 /(Adpn)2LiPF6 /Li 0 cell, current densities of 0.01mA/cm 2 for 1 to 120 cycles, 0.05mA/cm 2 for 121 to 180 cycles, and 0.1 mA/cm 2 for 181 to 240 cycles were used. Li + ion transference numbers were obtained by the method of combined ac and dc measurements 1,2 . The cathode was prepared from LiFePO4/carbon black/PVDF binder (8/1/1 by weight) using N-methyl-2-pyrrolidone (NMP)
to form a slurry that was doctor-bladed onto battery-grade aluminum foil to form 1.9-2.2 mg/cm 2 electrodes. The electrodes were dried in a vacuum oven overnight at 120°C. The dried electrodes were calendared with a Durston flat agile F130 mm rolling mill mechanical presser.
Force field parameters for bonded and Lennard Jones (vdW) interactions for Adpn were taken from the OPLS all-atom force field 3 . The partial charges on all atoms of Adpn were calculated from the MP2//aug-cc-PVDZ optimized structure using the CHELPG method 4 . Since the Li + ion is present in tetra coordination with Adpn, the partial charge on Li + was calculated using the optimized structure with a long range and dispersion corrected ωB97xD functional with 6-311++G(d,p)optimization/aug-cc-PVDZ charge calculation basis set. This calculation suggested a partial charge of 0.845 e -unit on the Li + ions. The charge value on the Li + ion (0.845e) was used as the scaling factor to rescale the partial charges on PF6 -ions (obtained from MP2//aug-cc-PVDZ calculations, separately). To compute theoretical Raman spectra for Adpn and (Adpn)2LiPF6, vibrational frequencies were calculated using PBE/6-311++G(d,p) for the structures optimized using the same functional/basis set. All the gas phase quantum chemistry calculations were carried out using the GAUSSIAN 9.0 software package 5 . Detailed protocol for force-field development is provided in Supplementary Information.
A supercell of 5x5x5 unit cells (20,000 atoms) of (Adpn)2LiPF6 was constructed in a cuboid with the dimensions of 55.35 x 64.75 x 63.25 Å 3 . This model was used for simulations under periodic boundary conditions to represent the bulk phase, and hence, is designated as model P. However, since the surface atoms have a large contribution towards the conduction and decomposition of these co-crystalline electrolytes, a different model V was used to understand the structure, dynamics and thermal behavior at the surface. Model V was constructed by placing a 5x5x5 supercell in a cube of 200 Å/side. GROMACS 5.0.7 software 6 was used for simulations and analysis along with VMD 1.9.3 software 7 for visualization of trajectories. The supercell models were energy minimized using standard protocols and algorithms implemented in the code. All the simulations for model P were carried out using NpT ensemble conditions, while for model V, canonical ensemble conditions were used. Details of temperature and pressure couplings and other MD parameters can be found in another work on a similar co-crystalline electrolyte material 8 .
To examine the mechanism of conduction at the atomic level with precise energetics, PW-DFT calculations were performed using the QUANTUM ESPRESSO 6.1 software package 9 . A projector-augmented-wave-basis set (Kresse-Joubert) 10 was used with PBE pseudopotentials 11 with a cut-off of 60 Ry and 360 Ry for kinetic energy and electron density, respectively. A unit cell of (Adpn)2LiPF6 cocrystals (from single-crystal XRD data) was relaxed in a fixed volume box and later in a variable cell manner to energy minimize the crystal structure. A threshold of 10 -7 Ry was used for electronic optimization and 10 -3 Ry/Bohr for force minimization. Table S3 shows a comparison of unit cell parameters obtained from single-crystal XRD data and variable cell relaxation DFT calculations. To investigate the path of Li + ion conduction in the cocrystal, 1x1x2, 3x1x1 and 2x1x2 supercells were created and optimized using the above discussed protocols. Due to the large system size of these supercells, all the calculations were performed with a Γ-only 1x1x1 k-mesh. For every supercell, a pair of Li + ion-defected configurations was used as an initial and final image for Nudged Elastic Band (NEB) calculations to interpolate the minimum energy path (MEP). Several sets of images were used to obtain the MEP, with a threshold of 0.1 eV/Bohr (f) pressed pellet that has been quenched in liquid N2 so that some grain boundaries broken; (g, h, i) powders, fusion between the grains can be observed by comparison between powders and powders pressed to make pellets; even in powder, there can be connection between the grains (i); in the pressed pellet the large grains contain smaller crystallites and there are needle-like structures between the grains; (j, k, l) samples synthesized in glass fiber. S2.