The Role of Electrolyte Composition in Enabling Li Metal‐Iron Fluoride Full‐Cell Batteries

Abstract FeF3 conversion cathodes, paired with Li metal, are promising for use in next‐generation secondary batteries and offer a remarkable theoretical energy density of 1947 Wh kg−1 compared to 690 Wh kg−1 for LiNi0.5Mn1.5O4; however, many successful studies on FeF3 cathodes are performed in cells with a large (>90‐fold) excess of Li that disguises the effects of tested variables on the anode and decreases the practical energy density of the battery. Herein, it is demonstrated that for full‐cell compatibility, the electrolyte must produce both a protective solid‐electrolyte interphase and cathode‐electrolyte interphase and that an electrolyte composed of 1:1.3:3 (m/m) LiFSI, 1,2‐dimethoxyethane, and 1,1,2,2‐tetrafluoroethyl‐2,2,3,3‐tetrafluoropropyl ether fulfills both these requirements. This work demonstrates the importance of verifying electrode level solutions on the full‐cell level when developing new battery chemistries and represents the first full cell demonstration of a Li/FeF3 cell, with both limited Li and high capacity FeF3 utilization.

S3 solution of 1M LiPF6 in a 1:1 (v/v) ratio of ethylene carbonate/dimethyl carbonate (EC/DMC) from Sigma Aldrich was used as received as the EC/DMC electrolyte.

Cathode Preparation
All cathode preparation was performed inside a glovebox under an Ar atmosphere. A slurry containing a 70:20:10 ratio of FeF3/carbon/binder in 1-methyl-2-pyrrolidinone (NMP, Sigma Aldrich, anhydrous) was prepared by stirring over 2 hours. Polyvinylidene fluoride (PVDF,5130 Solvay) was first dissolved completely in the NMP before dispersing the FeF3/C. Typical slurries had a solid content of 200 mg/mL of NMP. After mixing, the slurry was coated onto a piece of carbon-coated Al foil by doctor blading using a 20 μm slot. The film was dried at 65 °C overnight, then 1.98 cm 2 electrodes were punched from the substrate and used as cathodes in the batteries. The electrodes had an average areal loading of 2.44 mgFeF3/cm 2 and an average areal capacity of 1.74 mAh/cm 2 .

Full-Cell FeF3/C Battery Assembly
Please see Figure S20 for a detailed graphic of the coin cells described below and Table S1 for physical properties of the various cell components. All coin cells were constructed inside an Arfilled glovebox (MBraun) with H2O/O2 concentrations less than 0.2 ppm. For all electrolytes, CR2032 coin cells were constructed using a 1.98 cm 2 FeF3/C cathode and a 2.01 cm 2 Li anode; excess-Li cells used a 750 μm-thick anode (Alfa Aesar) and limited-Li cells used a 20 μm-thick anode on 10 μm Cu foil (Albermarle). In addition, a single wave spring and 0.7 mm of stainless steel spacer were also included inside the cell casing for 750 μm Li cells, while a single wave spring and 1.2 mm of stainless steel spacer were used for 20 μm Li cells. The same volume of electrolyte and number/type of spacer were used for both 750 μm and 20 μm Li cells. For cells containing Pyr13FSI electrolyte, 100 μL of electrolyte and a single 2.84 cm 2 glass fiber separator S4 (Whatman, GF/C) was used. For cells containing the TTE/DME and bisalt electrolytes, 70 μL of electrolyte and two 2.84 cm 2 polypropylene separator (Celgard 2400) were used. For the EC/DMC electrolyte, 70 μL electrolyte and two 2.84 cm 2 polymer separators (W-SCOPE) were used. Separators were chosen based on their wettability by each electrolyte; EC/DMC does not wet Celgard 2400 as well as W-SCOPE, while Pyr13FSI does not wet either polypropylene separator, instead requiring the use of glass fiber to allow for adequate wetting.

Half-cell Li/Cu Construction
Lithium-copper half-cells were prepared for Li plating Coulombic efficiency measurements.
Unless otherwise specified, all materials were used as received. Cu working electrodes (9 μm) were treated with a 1.2 M HCl bath to remove the native oxide. Following the HCl bath the Cu electrodes were washed with 18 MΩ water and then acetone. The Li counter electrodes, unless otherwise noted, were a 50 μm Li metal laminate on 10 μm of Cu foil (Albermarle). 750 μm Li (Alfa Aesar), used in specified tests, was also used. Electrodes were punched into 16 mm discs and had an area of 2.01 cm 2 . The coin cells prepared with the TTE/DME and bisalt electrolytes used two Celgard 2400 separators with 80 μL of electrolyte and the coin cells prepared with EC/DMC used two W-SCOPE separators with 80 μL of electrolyte; these cells all used the 50 μm-thick Li. The coin cells prepared with the Pyr 13 FSI electrolyte used one glass fiber separator (Whatman, GF/C) and 100 μL of electrolyte and the 750 μm-thick Li. As previously mentioned, the choice for each separator was based on the wettability of each separator material with each electrolyte. The use of the 750 μm-thick Li with the ionic liquid electrolyte was based on supplemental lithium cycling measurements as discussed below.

S5
Powder X-ray diffraction (XRD) spectroscopy was performed on a D2 Phaser (Bruker) spectrometer using a Cu kα radiation source. Cells used for physical characterization were cycled twice and stopped at a charged state (4.0V) before being opened inside the Ar-filled glovebox to harvest cathodes and anodes. The electrodes were soaked in DME for 5 seconds to remove any residual electrolyte from the surface and allowed to dry prior to further processing. Scanning electron microscopy (SEM) images were collected using a Supra 55VP Field Effect Scanning Electron Microscopy (Zeiss) at 3 kV accelerating voltage using both an electron backscatter detector (EBSD) and a secondary electron detector. All samples were loaded into the microscope using an Ar-filled glovebag to limit exposure of the samples to oxygen and moisture. X-ray photoelectron spectroscopy (XPS) was performed using a K-alpha X-ray photoelectron spectrometer (Thermo Scientific) using a monochromatic Al source (Kα = 1486.6 eV). Samples were loaded into the XPS using an inert atmosphere transfer arm to prevent exposure to moisture and oxygen. All spectra were corrected to an adventitious (sp3) carbon peak at 284.8 eV.
Scanning transmission electron microscopy (STEM) images were captured using a Titan G2 80-200 microscope (FEI Company) operated at 200 kV and equipped with four silicon-drift X-ray detectors (Super X TM ). Electron energy loss spectra (EELS) were collected using a Quantum 963 spectrometer (Gatan, Inc.). Samples for STEM analysis were prepared by sonicating a small piece of the cathode film in dry ethanol inside a sealed bottle to suspend some of the FeF3/C composite, then cast onto lacey carbon grid and allowed to dry inside the Ar glovebox. They were then transferred into the STEM instrument inside an inert atmosphere transfer arm to limit exposure to ambient atmosphere and moisture.

Electrochemical Characterization
Full-cell FeF3/Li cells and limited Li cells were tested galvanostatically on a Series 4000 battery tester (Maccor). Except where noted, cells were cycled between 1 and 4 V at a C/20 rate relative to the capacity of the FeF3 in the cathode. Average cell capacities were calculated from triplicate measurements of identically constructed and cycled cells using each electrolyte, and were reported with error bars equivalent to a single standard deviation from the mean. For FeF3/Li cells, Coulombic efficiency is defined according to Equation S1. Equation S1: The lithium Coulombic efficiency measurements were made using "Method 3" as described by Adams, et al. in their previous work using Arbin battery cyclers. 2 First, a lithium reservoir of 4 mAh/cm 2 was plated at 0.1 mA/cm 2 , next the reservoir was stripped, and replated. Then 0.5 mAh/cm 2 of charge was cycled 51 times before the final, exhaustive stripping step. A +/-1 V voltage limit was applied for the stripping/plating steps. The average Coulombic efficiency was calculated according to Equation S2,as described in the literature 3 :

Equation S2
: where n is the number of cycles, QT is the amount of charge plated in the lithium reservoir, Qc was the amount of charge passed through over the 51 cycles, and Qs was the final amount of charge stripped. For cells that polarized (and therefore could not deliver the designated amount of charge) before reaching the 51 cycles, the total amount of plating and stripping charge were summed over S7 the 51 cycles and that value replaced the nQc term in the equation. The charge from the first, or formation cycle, is not included in this calculation.
Some supplementary lithium/copper cycling tests were also completed with the Pyr 13 FSI electrolyte. For these tests, 0.5 mAh/cm 2 of lithium was plated at 0.1 mA/cm 2 (or until it reached a -1 V voltage cutoff) and then lithium was stripped until it reached a 1 V limit. A maximum time limit, resulting in twice the amount of charge (1 mAh/cm 2 ), was used on the stripping step to identify soft short circuits or parasitic reactions that artificially inflate the Coulombic efficiency measurements.
Cells used for SEM, STEM, and XPS analysis were discharged and charged twice at C/20 prior to being opened inside an Ar glovebox. The anode and cathode were harvested from the opened cells and dunked in DME for 5 seconds to remove any residual electrolyte solution and left to dry.

Statistical Analysis
1) Pre-processing of data: Capacities measured from cycling data was transformed from units of "mAh" reported by the instrument to "mAh/g" by dividing this capacity by the mass of the FeF3 in the cell's cathode, for the excess-Li cells, or by the total mass of the cathode and the electrolyte together, for the limited-Li cells. For XPS measurements, a Shirley background was used to fit all of the spectra and subsequently subtracted before plotting. 2) Data presentation: All graphs presenting the averaged results of electrochemical tests (e.g. Figure S1) are presented as the mean +/-standard deviation (SD), with error bars or shaded areas representing the standard deviation. 3) Sample size: All electrochemical tests were performed in triplicate (n = 3); all physical characterizations were performed on a single sample (n = 1). 4) Statistical methods: No S8 statistical differentiation was performed in this work. 5) Software: Data was processed in Microsoft Excel, Origin Labs Origin 2020, and KasaXPS.

Supplementary Discussion: Coulombic Efficiency Measurements using Pyr13FSI
Supplemental Li plating and stripping measurements using Pyr13FSI electrolyte were also conducted (Supplemental Discussion Figure 1). For these measurements, Li was plated and stripped on a Cu working electrode at 0.1 mA/cm 2 for 0.5 mAh/cm 2 . Both thin (50 μm) and thick (750 μm) Li were cycled under these parameters to determine the Coulombic efficiency of Li cycling. We found that the cell using the 750 μm Li is able to cycle reasonably for the first 10 cycles but the cells with 50 μm Li had Coulombic efficiencies greater than 100 % as early as the first cycle. This suggests that Li is not able to effectively cycle using the thinner 50 μm Li with the Pyr 13 FSI electrolyte, and therefore the 750 μm Li was used for the average Coulombic efficiency measurements. Even still, the average Coulombic efficiency measurements of the Pyr13FSI electrolyte resulted in significant variability and unreliable measurements, exemplified by greater than 100 % CE on the formation cycle with 90 % average CE, as well as an average calculated CE that is greater than 100 % efficiency (see Figure S7 for replicates). Furthermore, the overpotential was much lower than that of the other electrolytes (around 3 mV versus around 30 mV overpotentials for lithium plating). While this low overpotential could indicate that plating and stripping is less kinetically limited in the Pyr13FSI electrolyte, it more likely suggests soft shorts, given the other data that points to difficulty cycling. The cause of the sporadic behavior with the 750 μm Li may be due to parasitic reactions during cycling or Li growth S9 through the glass fiber separator leading to partial shorts within the cell. Ultimately it was determined that the Pyr 13 FSI electrolyte can cycle Li metal with moderately low efficiencies (80-91 %) for a limited number of cycles but failure is impending.     TTE/DME, c) bisalt, and d) EC/DMC coin cells cycled at C/20. All cells show a distinct twophase initial discharge with plateaus at approximately 3.1 V and 1.6 V; this second plateau moves to 2.25 V in subsequent discharges. The initial charge shows two plateaus and appears similar to subsequent charge steps, though a third plateau gradually appears around 2.25 V after cycle 10.  Figure S8: Rate-cycling comparison of Pyr 13 FSI and TTE/DME cells versus 750 μm Li foil (n = 1). TTE/DME is able to cycle with higher capacity at faster rates (1C) than Pyr 13 FSI and recovers its initial capacity upon return to slower rates (C/20).       Figure S19: a) XPS survey spectra and spectra from the b) C1s, c) Fe 2p, d) F1s, e) O1s, f) S 2p, and g) N 1s regions of fresh and tested (two cycles) FeF 3 cathodes with assigned peaks.   Figure S22: Average discharge of FeF 3 /Li cells, tested using limited-Li metal anodes (20 μm) in Pyr 13 FSI, TTE/DME, bisalt, and EC/DMC electrolytes at C/20. Capacity is normalized per gram of FeF 3 in the cathodes. Averages and standard deviation (shaded areas) are calculated from three cells (n = 3) and reported as mean±SD. Figure S23: Schematic of a typical coin cell. a) stainless steel anode cap, b) stainless steel wave spring, c) 0.1 mm thick x 15.5 mm diameter stainless steel spacer, d) 0.5 mm thick x 16 mm diameter stainless steel spacer, e) 0.75 mm thick x 16 mm diameter Li foil, f) 19 mm diameter separators, g) 16 mm diameter FeF 3 /C/PVDF cathode on C-coated Al foil, h) stainless steel cathode cap. Two Celgard 2400 separators are used to build the TTE/DME and bisalt electrolyte cells, two W-SCOPE separators are used for the EC/DMC electrolyte, and one Whatman glass microfiber (GF/C) separator was used with Pyr 13 FSI electrolyte. 70 μL of electrolyte was used for the TTE/DME, bisalt, and EC/DMC electrolytes, while 100 μL of the Pyr 13 FSI electrolyte was used.