Opening Diffusion Pathways through Site Disorder: The Interplay of Local Structure and Ion Dynamics in the Solid Electrolyte Li6+xP1–xGexS5I as Probed by Neutron Diffraction and NMR

Solid electrolytes are at the heart of future energy storage systems. Li-bearing argyrodites are frontrunners in terms of Li+ ion conductivity. Although many studies have investigated the effect of elemental substitution on ionic conductivity, we still do not fully understand the various origins leading to improved ion dynamics. Here, Li6+xP1–xGexS5I served as an application-oriented model system to study the effect of cation substitution (P5+ vs Ge4+) on Li+ ion dynamics. While Li6PS5I is a rather poor ionic conductor (10–6 S cm–1, 298 K), the Ge-containing samples show specific conductivities on the order of 10–2 S cm–1 (330 K). Replacing P5+ with Ge4+ not only causes S2–/I– anion site disorder but also reveals via neutron diffraction that the Li+ ions do occupy several originally empty sites between the Li rich cages in the argyrodite framework. Here, we used 7Li and 31P NMR to show that this Li+ site disorder has a tremendous effect on both local ion dynamics and long-range Li+ transport. For the Ge-rich samples, NMR revealed several new Li+ exchange processes, which are to be characterized by rather low activation barriers (0.1–0.3 eV). Consequently, in samples with high Ge-contents, the Li+ ions have access to an interconnected network of pathways allowing for rapid exchange processes between the Li cages. By (i) relating the changes of the crystal structure and (ii) measuring the dynamic features as a function of length scale, we were able to rationalize the microscopic origins of fast, long-range ion transport in this class of electrolytes.

. Exemplary Rietveld refinements against neutron powder diffraction data measured at 150 K and 200 K relative to Li6+xP1-xGexS5I (x = 0.25, 0.6) and Li6PS5I, respectively. Experimental data are shown in black and the red line denotes the calculated pattern, whereas the difference profile is shown in blue. Positions of the argyrodite Bragg reflections are indicated by green vertical ticks. The low values for the fit indicators together with the flat profile difference confirm the good quality of the calculated structural model.

S2
Besides stabilizing the cubic phase, the Ge substitution expands the unit cell volume, due to the larger ionic radius of Ge 4+ compared to P 5+ . 1 The lattice parameters obtained from Rietveld refinements of the three selected compositions for the low-temperature neutron studies (x = 0.0, 0.25, 0.6) are shown in Figure S2a, together with the lattice parameters obtained from profile fit of room-temperature X-ray data for all other studied compositions. Due to the lower temperature at which the diffraction experiments were performed, the lattice parameters carried out from neutron data are systematically smaller than the ones obtained from the roomtemperature X-ray data. The linear increase of the lattice parameters upon substitution of Ge 4+ by P 5+ indicates the successful synthesis of stable solid solutions.
Further insights into the structural modifications brought up by the elemental substitution can be provided by Rietveld refinements. To corroborate the incorporation of the substituent, the occupancy of Ge on the nominal P site (Wyckoff 4b) was allowed to refine and the obtained values are reported against the nominal content in Figure S2b. The linear correlation between refined and nominal occupancy indicate the successful substitution that is further supported by the volume expansion of the (P1 -xGex)S4 tetrahedra upon Ge substitution (see Figure  S2c). Figure S2. The linear increase of (a) the lattice parameters, carried out from (gray) X-ray and (blue) neutron data, together with (b) the refined Ge occupancy against the nominal Ge content xn indicate the formation of a true solid solution. Moreover, (c) the expansion of the (P/Ge)S4 units against the refined Ge content xr corroborates the successful substitution of Ge for P in such tetrahedral environment.   Figure S4. 6 Li MAS NMR spectra of Li6+xP1-xGexS5I with different Ge-contents. Note the asymmetric shape of the 6 Li spectra pointing to several contributions to the signal.
With higher Ge-content the 6 Li lines measured under MAS conditions show a slight upfield shift, i.e., to lower ppmvalues. A similar trend is also observed for the 31 P MAS lines. It results from the fact that coupling of the respective nuclei with less electronegative partners becomes more dominant. As Ge 4+ exchanges P 5+ upon substitution and also additional Li + ions are incorporated, the chemical environment of the 6 Li species shifts upfield, resulting also in a small change in chemical shift in the respective spectra.  Polarisation measurements. For the dc polarisation measurement, a pressed pellet (d = 8 mm; h=0.88mm (20 at%) 1.14 mm(60 at%) was sputtered with ion-blocking Au electrodes on both sides. The sample preparation was carried out under Ar atmosphere. We measured the pellet in an air-tight 2-electrode Swagelok-type cell connected to a Parstat MC potentiostat (Princeton Applied Research) equipped with a low-current option. The chronoamperometric polarization experiments were performed in a Faraday cage at ambient temperature for 7 days (604800 s, 10 s/point) with a potential set to 0.1 V.
For both samples the overall conductivity σ reaches a limit in the order of 10 −9 Scm −1 which is due to the electronic contribution to the total conductivity. This value represents the upper limit of any electronic conductivity in the samples investigated. Here, σeon seems to be almost independent of the Ge-content in the investigated samples. However, considering the large difference in total conductivity between the samples with 10 at% and 60 at% Ge, the electronic contribution, that is, the respective transference number, is higher in the poor ionic conductor with 10 at% Ge. Figure S7. Chronoamperometric evolution of the overall conductivity over 7 days for for the samples Li6+xP1-xGexS5I with (a) 10 at% Ge and (b) 60 at% Ge. The applied potential was set to 0.1 V. Both samples reach a limiting value at sufficiently long waiting time, which is identified to be governed only by electronic conductivity σeon. Here, σeon turned out to be in the order of 10 -9 S cm -1 or lower.