Dynamics of Solid‐Electrolyte Interphase Formation on Silicon Electrodes Revealed by Combinatorial Electrochemical Screening

Abstract Revealing how formation protocols influence the properties of the solid‐electrolyte interphase (SEI) on Si electrodes is key to developing the next generation of Li‐ion batteries. SEI understanding is, however, limited by the low‐throughput nature of conventional characterisation techniques. Herein, correlative scanning electrochemical cell microscopy (SECCM) and shell‐isolated nanoparticles for enhanced Raman spectroscopy (SHINERS) are used for combinatorial screening of the SEI formation under a broad experimental space (20 sets of different conditions with several repeats). This novel approach reveals the heterogeneous nature and dynamics of the SEI electrochemical properties and chemical composition on Si electrodes, which evolve in a characteristic manner as a function of cycle number. Correlative SECCM/SHINERS has the potential to screen thousands of candidate experiments on a variety of battery materials to accelerate the optimization of SEI formation methods, a key bottleneck in battery manufacturing.

S3 of predefined locations on the Si surface through an automated hopping regime consisting of pipet approach to the surface at a rate of 2 µm s -1 , electrochemical measurement, retract at 2 µm s -1 , and x-y translation at 2 µm s -1 , as previously reported. [3,4] The lateral separation between each individual measurement was set to 12 or 15 µm to prevent spatial overlap of the probed areas.
SECCM was deployed under a combinatorial protocol to evaluate the SEI formation under a broad experimental space by translating the pipet probe to another location and repeating the measurement protocol under a different set of conditions. This approach involved the combination of two different cutoff voltages (+0.05 V and -0.13 V vs Li/Li + , Figure S2) and a different number of charge/discharge cycles (1, 2, 5, 10, 15 cycles). The same protocol was repeated using a new pipet probe filled with a different electrolyte (1 M LiPF6 in EC/EMC or 1 M LiPF6 in PC). Full combinatorial space is illustrated in Figure S3 and the footprint left by SECCM measurements is shown in Figure S4. Current densities are given normalised by the geometric area of the SECCM footprints. Note that to facilitate the finding of SECCM footprints and avoid any possible dissolution of SEI components and compositional changes, the Si samples were not rinsed after SECCM to carry out the SHINERS analysis. As a consequence, they contain a small amount of residual electrolyte salts. Samples were kept under inert atmosphere during transfer between instruments.
The SECCM setup (sample and positioner) was placed on a bench-top vibration isolation platform (BM-10, Minus K Technology) to minimize mechanical vibrations and covered with a copper woven mesh (60 mesh per inch, 0.166 mm wire diameter, Cadisch Precision Meshes) acting as Faraday cage to reduce electrical noise. Data acquisition and instrumental control in SECCM was achieved using a FPGA card (PCIe-7852R) controlled by a LabVIEW 2020 (National Instruments) interface running the Warwick electrochemical scanning probe microscopy (WEC-SPM, www.warwick.ac.uk/electrochemistry) software.
Data processing and analysis was carried out with a code written in-house in Python language with the aid of SciPy libraries. [5] Microscopic characterisation of the SEI after SECCM The morphology and thickness of the SEI was analysed by atomic force microscopy (AFM) and SEM, whereas elemental mapping was recorded by energy-dispersive X-ray spectroscopy (EDX). AFM was carried out using an Innova microscope (Bruker) in tapping mode with antimony (n) doped Si probes (RFESP-75, Bruker). Scans were recorded with 256 points per line at 0.1 Hz. AFM images were analysed with the Gwyddion software (Czech Metrology Institute). Note that for the microscopic characterisation of the SEI, the electrolyte residues from the SECCM experiment were removed by gently dipping the Si sample in dimethyl carbonate (DMC) for a few seconds. While this is necessary to characterise the actual thickness and morphology of the SEI, without any interference from electrolyte residues, we acknowledge that some soluble SEI components might be dissolved through this process (vide infra); our goal was to obtain semiquantitative insight and trends of SEI thickness as a function of cycle number.

Synthesis of gold nanoparticles
Citrate-stabilised gold nanoparticles (Au NPs) were obtained by the citrate reduction method described by Turkevich and Frens [6][7][8][9] where 200 mL of a 0.01 % HAuCl4 (99.995 %, Sigma-Aldrich) solution was left to boil under strong stirring. 1.5 mL of 1 % citric acid (≥ 99.5 %, Sigma-Aldrich) were immediately added to the solution of HAuCl4 and the solution changed from pale yellow to maroon after a few minutes. The dispersion of nanoparticles was then allowed to cool under stirring and the resultant suspension was stored away from light sources for the duration of the experiments, remaining stable throughout the process.

Synthesis of SiO2-coated shell isolated nanoparticles (SHINs)
SiO2-coatings for SHINs were obtained following the protocol described by Tian et al. [10]  sodium silicate solution (Na2SiO3) (≥10% NaOH, ≥10% SiO2 basis, Sigma-Aldrich) were added dropwise and allowed to stir for another 3 minutes before transferring the flask to a hot water bath at 98°C. Suitable, pinhole-free SHINs were obtained after 30 minutes reaction time.

Characterisation of Au NPs and SHINs
Au NPs were characterised by UV-Visible absorption spectroscopy ( Figure S5). The position of the absorption band was used to determine the average size of the Au NPs. Au NPs had an estimated diameter of 68 nm, as determined by the method proposed by Haiss [11] and co-workers, for particles in the range of 35-110 nm diameter.
Differential centrifugal sedimentation (DCS) was carried out to investigate the SiO2 coating process of AuNPs ( Figure S6). Au NPs had a diameter of 57 ± 9.2x10 -3 nm according to DCS with a resolution of ±0.1 nm. [12] SiO2 shell thicknesses of 2, 4 and 8 nm were estimated for 15, 30 and 60 minutes respectively, following the iterative fitting described by the literature. [12,13] A disc centrifuge (DC24000, CPS Instruments Inc.) operated at 24000 rpm was used for DCS. Spin gradient was created by progressive addition of 8 and 24 wt % fresh sucrose (≥ 99.5%, Sigma-Aldrich) solutions. Calibration of the equipment was achieved before each measurement with polyvinyl chloride (PVC) particles (~0.263 μm, Analytik Ltd) as a standard.
All measurements were performed at least three times to ensure the reproducibility of the data.
Pinhole and enhancement tests were performed to ensure the suitability of the SiO2-coated Au NPs for SHINERS. [14] No pinholes were observed for SHINs deposited on Si wafer while a large enhancement was observed for SHINs deposited onto Au wafer ( Figure S7). These tests were carried out with a Raman microscope (Renishaw InVia) using a 633 nm laser at 10 %. Si wafer was used to calibrate the Raman with a resolution of 1.1 cm -1 . For Raman pinhole tests, 5 µL of concentrated SiO2-coated SHINs (~ 1.13 nM) were deposited onto a silicon wafer (Si (100), Agar Scientific) and 2 µL of a 10 mM pyridine (99.8 %,

S5
Sigma-Aldrich) solution were added on top of the SHINs. Enhancement tests were similarly performed using a gold wafer (Au (111), Platypus Technologies).

Co-located Shell-isolated nanoparticle-enhanced Raman spectroscopy (SHINERS)
Co-located SHINERS was performed on Si samples following combinatorial SECCM to analyse the chemical composition of the SEI layer. SHINs were dispersed in MilliQ water, and a drop deposited on a glass window. After drying in vacuum at 100 °C (overnight), the glass window with dispersed SHINs was placed in close contact with the SECCM Si sample on top of the already formed SEI ( Figure S8). These ex situ cells were assembled under argon atmosphere in glovebox with 0.1 ppm H2O and O2 levels.
A Raman microscope (Renishaw InVia) with a 633 nm laser as excitation source (power < 300 µW) was then used to record the Raman measurements. The laser was focused onto individual SECCM spots with a 50x objective. Raman spectra were collected with 5 accumulations and 30 s exposure time for each SECCM spot. Conditions were optimized to ensure good signal-to-noise ratio to resolve Raman bands.
Raman spectra were baseline-corrected using the baseline correction tool of the WiRE 4.0 software (Renishaw). A series of Raman spectra recorded for a set of SECCM repeats under the same conditions are shown in Figure S9. The spectra are variable depending on spot, which could be a result of SHINs density, heterogeneous enhancement and/or heterogeneous nature of SEI. Figure S1. SEM image of a typical single-channel pipet probe used for the SECCM measurements. The tip diameter was ca. 1 µm. Images were taken at an acceleration voltage of 5 kV using the InLens detector.     of 57 with a standard deviation of 9.2x10 -3 nm according to DCS with a resolution of ±0.1 nm. [12] Shell thicknesses of 2, 4 and 8 nm were estimated for 15, 30 and 60 minutes respectively, following the iterative fitting described by the literature. [12,13] Figure S4).              Peak (cm -1 ) Assignment Ref. 300 Au-SiO2 SHIN background -520 Si [19] 839

1341
ROCO2Li us(C O) [20,22] 1367 Au-SiO2 SHIN background -1445 dCH2 (EC, EMC and PC) + LMC [21] 1485 LMC [21] 1550 Bridging + monodentate carboxylate [17] 1580 Monodentate carboxylate RCOOLi uas(C O) [20,22] 1628 Monodentate carboxylate ROCO2Li uas(C=O) [20,22] 1687 ROCO2Li uas(C=O) [20,22] Complex Raman spectra are measured in-line with expectations that the SEI is an evolving structure made up of a range of chemical species in differing environments. This complexity makes definitive assignments to individual species impractical. Previous studies have identified ROCO2Li and RCOOLi as products of PC decomposition [20,23] and modes here around 1300 cm -1 and 1500-1700 cm -1 could be reasonably assigned to these species. Evidence for PC being present at the surface is through the presence of known modes at ca. 940 and 1140 cm -1 . [21] Similarly experiments using EC/EMC are in-line with formation of complex mixtures containing species such as RCOOLi or ROCO2Li (ca. 1340 cm -1 , 1500-1600 cm -1 ). [20,22,23] In some spectra bands pertaining directly to lithium methyl carbonate (LMC) can also be tentatively assigned. In all spectra the strong Raman mode at 520 cm -1 is due to the Si itself.