Experimental

Preparation of LPS samples.—LPS electrolyte was prepared at Solid Power Inc. using two routes, a previously reported procedure23 for X-ray CT experiments and ball milling for EDS analysis. The LPS powder was subsequently tape cast, and then transferred to 35-μm-thick lithium foil (RockwoodLithium,0.5%Al )by cold-pressing to form LPS/Li bilayers. Symmetric Li/LPS/Li structures were assembled using 3-mm discs of the Li/LPS bilayer and 2-mm discs of 35-μm-thick lithium foil, stacked and pressed in a coin cell crimper (Xiamen Tmaxcn Inc.) to ensure contact. Inside a vacuum glove box (Xiamen Tmaxcn Inc.), symmetric cells were placed into airtight in-situ cells (Fig. 1a) then removed from the glove box and biased for 1 hour with a current density of 100 μA/cm2 before the direction of the current was reversed for 1 hour. Current densities were calculated with respect to the area of the smaller, 2-mm-diameter lithium electrode. This process was repeated up to five times. Quasi-in-situ cells were cycled using a BioLogic VMP3 potentiostat and in-operando cells were cycled using a Keithley Sourcemeter Model 2450. All cycling and analysis was conducted at room temperature.

Design of in-situ cell.—In this work, a specialized sample holder was designed (Fig. 1) to allow for tomographic characterization of symmetric Li metal/LPS/Li metal structures at relatively low synchrotron X-ray energies and under external bias using designs reported in the literature29 as a starting point and further optimized for these experiments. The design was tailored to maximize visibility of lithium metal, which is not strongly X-ray-absorbing, in samples with this specific geometry. The sample holder maintained electrical contact with the lithium metal electrodes to enable in-situ and in-operando X-ray CT studies. The brass current collectors were interfaced with carbon-based current collectors to avoid X-ray absorption by the sample holder near the region of interest, potentially introducing extraneous interactions to the electrochemical system, which will be discussed briefly in the Results section. However, our work focused on determining the necessary geometry and analysis conditions to extract in-operando tomographic data. The plastic housing used for the quasi-in-situ experiments was Delrin acetal resin (10% PTFE), machined to a 0.5-mm thickness at the region of interest. The plastic housing used for the in-operando experiments was 2.5-mm-thick Torlon polyamideimide to provide an increased resistance to radiation damage over acetal. A small spring test probe ensured electrical contact in case of sample-thickness variation and applied a small amount of pressure during cycling. The housing was sealed to the metal current collectors using Swagelok compression fittings and the assembly was leak-tested prior to loading samples.