Lithium Self-Discharge and Its Prevention: Direct Visualization through

Nov 7, 2017 - To understand the mechanism that controls low-aspect-ratio lithium deposition morphologies for Li-metal anodes in batteries, we conducte...
0 downloads 9 Views 9MB Size
Download PDF
Subscriber access provided by READING UNIV

Article

Lithium Self-Discharge and its Prevention: Direct Visualization through In-Situ Electrochemical Scanning Transmission Electron Microscopy Katharine L. Harrison, Kevin R. Zavadil, Nathan T. Hahn, Xiangbo Meng, Jeffrey W. Elam, Andrew J. Leenheer, Ji-Guang Zhang, and Katherine L. Jungjohann ACS Nano, Just Accepted Manuscript • DOI: 10.1021/acsnano.7b05513 • Publication Date (Web): 07 Nov 2017 Downloaded from http://pubs.acs.org on November 9, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

ACS Nano is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

Lithium Self-Discharge and its Prevention: Direct Visualization through In-Situ Electrochemical Scanning Transmission Electron Microscopy Katharine L. Harrison,†ǂ Kevin R. Zavadil,†ǂ Nathan T. Hahn,†ǂ Xiangbo Meng,†ǁ Jeffrey W. Elam,†ǁ Andrew Leenheer,┴ Ji-Guang Zhang,†# and Katherine L. Jungjohann†§* † ǂ

Joint Center for Energy Storage Research

Nanoscale Sciences, ┴ MESA Fabrication Operations, § Center for Integrated Nanotechnologies, Sandia National Laboratories, Albuquerque, New Mexico 87123, USA ǁ #

Argonne National Laboratory, Lemont, Illinois 60439, USA

Energy & Environmental Directorate, Pacific Northwest National Laboratory, Richland, Washington 99352, USA

KEYWORDS: electrochemical transmission electron microscopy, lithium-ion batteries, solid electrolyte interphase, mechanical compression, protective coating, artificial solid electrolyte interphase, lithium metal anode ABSTRACT To understand the mechanism that controls low-aspect-ratio lithium deposition morphologies for Li metal anodes in batteries, we conducted direct visualization of Li metal deposition and stripping behavior through nanoscale in-situ electrochemical scanning transmission electron microscopy and macroscale-cell electrochemistry experiments in a recently-developed and 1 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 42

promising solvate electrolyte, 4 M lithium bis(fluorosulfonyl)imide in 1,2-dimethoxyethane. In contrast to published coin cell studies in the same electrolyte, our experiments revealed low Coulombic efficiencies and inhomogeneous Li morphology during in-situ observation. We conclude that this discrepancy in Coulombic efficiency and morphology of the Li deposits was dependent on the presence of a compressed lithium-separator interface, as we have confirmed through macroscale (not in the transmission electron microscope) electrochemical experiments. Our data suggests that cell compression changed how the solid electrolyte interphase formed, which is likely responsible for improved morphology and Coulombic efficiency with compression. Furthermore, during the in-situ electrochemical scanning transmission electron microscopy experiments, we observed direct evidence of nanoscale self-discharge in the solvate electrolyte (in the state of electrical isolation). This self-discharge was duplicated in the macroscale, but it was less severe with electrode compression, likely due to a more passivating and corrosion-resistant solid electrolyte interphase formed in the presence of compression. By combining the solvate electrolyte with a protective LiAl0.3S coating, we show that the Li nucleation density increased during deposition leading to improved morphological uniformity. Furthermore, self-discharge was suppressed during rest periods in the cycling profile with coatings present, as evidenced through EC-STEM and confirmed with coin cell data. Lithium-ion batteries have become the ubiquitous energy storage solution for mobile power applications ranging from portable electronics to electric vehicles. However, improvements in Li-ion battery technology are necessary to keep pace with the increasing energy and power demands of select applications. Many strategies have been used to improve Li-ion batteries including alternative cathode, anode, and electrolyte materials, but the largest improvements would result from using Li-metal anodes rather than intercalation or alloy 2 ACS Paragon Plus Environment

Page 3 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

materials.1-3 Furthermore, the promise of next-generation battery concepts, that venture beyond Li-ion, such as Li-S and Li-air batteries, depend on the successful implementation of safe, longcycle life, and high Coulombic efficiency (ηCoulombic) Li-metal anodes to achieve their proposed energy densities.2-4 Therefore, a mechanistic understanding of high efficiency Li-metal anode deposition and stripping processes is critical to the development of higher energy density storage solutions. Li-metal anodes have been studied for decades, but the challenge of producing stable, long cycle life, high efficiency Li electrodes within a liquid-electrolyte cell remains.1-9 Li deposition and stripping is likely to be particularly difficult to implement in full cells with intercalation cathodes, as opposed to the half cells that are typically studied, due to factors such as volume expansion mismatch. Because Li metal is very reactive and deposits at a very low potential (-3.04 V vs. SHE), electrolytes are generally unstable in the presence of Li. As a result of this instability, a solid electrolyte interphase (SEI) grows during Li deposition, which beneficially limits further reaction between the anode and the electrolyte. However, the formation of the SEI decreases the ηCoulombic, increases impedance, and depletes available Li and electrolyte within the cell chemical inventory. Another problem with Li metal is that it tends to deposit in high-surface-area, non-uniform morphologies (mossy, high-aspect-ratio dendrites, etc.) which leads to large quantities of SEI growth. SEI irregularities are also often implicated in causing non-uniform Li morphology. The non-uniform Li morphology allows Li to be easily stranded during stripping, leading to loss of electrical connection to the current collector (“dead Li”) during cycling. The dead Li further depletes the Li supply in the cell and decreases ηCoulombic. In addition to performance degradation, the non-uniform Li morphology also constitutes a serious safety problem: high-aspect-ratio structures can propagate across the cell 3 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 42

and penetrate through the separator, leading to cell shorting, excess heat generation (fires), and catastrophic cell failure (explosions). This is a critical issue for many user applications such as travel with portable electronics on airplanes and in consideration of portable batteries used in medical devices within the body, such as pacemakers. Many strategies have been employed to promote uniform, stable Li-metal deposition and stripping, which in turn can improve performance and safety.8-11 Redesign of the electrolyte,12,13 along with the use of additives, has proven an effective method for favorably impacting SEI formation, ηCoulombic, and growth morphology. Additionally, the concept of an artificial SEI has been explored by film deposition at the anode14-17 to regulate the electrolyte reaction with the evolving Li surface. We initially aimed to investigate these two methods (a recently demonstrated high-performance designer electrolyte and a protective artificial SEI layer) at the nanoscale to evaluate their influence on the stability and morphological evolution of the Li electrode-electrolyte interface. We targeted a solvate electrolyte composed of 4 M lithium bis(fluorosulfonyl)imide in 1,2-dimethoxyethane (4M LiFSI/DME) that was demonstrated by Qian et al. in 2015.12 This work showed that 4M LiFSI/DME provided high ηCoulombic over long term cycling on Cu at relevant current densities in coin cells (>97% at 10 mA/cm2 for >500 cycles, 98.4% at 4 mA/cm2 for 1000 cycles, and 99.1% at 0.2 mA/cm2 for 500 cycles).12 The high ηCoulombic was attributed to low reactivity of the highly concentrated electrolyte with Li metal, leading to a compact SEI film. In addition, the morphology of Li electrodes cycled in coin cells was shown to be very uniform even after long-term cycling, forming dense columnar grains without evidence of highaspect-ratio dendritic structures.12 Therefore, we chose this electrolyte to perform nanoscale observation of the Li deposition/stripping behavior to identify the key mechanistic difference 4 ACS Paragon Plus Environment

Page 5 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

controlling the Li morphology as compared to high-aspect-ratio dendritic Li structures that are known to abundantly form in carbonate-based electrolytes. We additionally explored the possibility of improved performance of Li-metal anodes in this electrolyte in the presence of a 50 nm atomic layer deposition (ALD) film of LiAl0.3S.14 LiAl0.3S coatings have been previously shown to promote denser, flatter Li deposits, suppress high-aspect-ratio dendrites, and improve ηCoulombic over long term cycling both in carbonate- and ether-based electrolytes.14 Furthermore, the impedance of cells containing LixAlyS-coated Li electrodes have been shown to be much lower than that of cells without LixAlyS.14 Finally, the lower impedance has been correlated to a reduced extent of reaction between the Li and electrolyte and therefore minimal SEI formation. In this work, we performed in-situ electrochemical scanning transmission electron microscopy (EC-STEM) to directly visualize Li deposition and stripping in 4M LiFSI/DME with and without ALD LiAl0.3S protective coatings. Unexpectedly, with the solvate electrolyte, nonuniform morphologies were observed in the EC-STEM images. The morphology resembled what has been previously observed at the nanoscale for carbonate-based electrolytes18 despite demonstrated coin cell data showing uniform columnar grain growth during Li deposition in 4M LiFSI/DME.12 Through comparisons between macroscale electrochemical experiments in a coin cell with compression (and a separator) vs. a beaker cell (without compression), we found that the absence of compression led to non-uniform morphology and low ηCoulombic in 4M LiFSI/DME. We provide evidence that the SEI formed differently on the Li electrode in the absence of compression, which resulted in poor performance. Compression has been previously shown to affect Li deposition and stripping in a few other electrolytes.4,

19-22

However, we

demonstrate that even in the absence of compression, increased nucleation density and improved Li morphology can be achieved in nanoscale and macroscale cells with the use of ALD LiAl0.3S 5 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 42

protective coatings. Furthermore, we observed direct, visual evidence of nanoscale self-discharge during open circuit rest after electrodeposition of Li in the EC-STEM experiments. Selfdischarge was confirmed in coin cells, but was observed to a lesser extent in these cells, which were cycled under compression. Self-discharge was further suppressed in cells with LiAl0.3Scoated current collectors, as evidenced through both EC-STEM results and coin cell electrochemical data. In summary, this work shows that: (a) cell compression in 4M LiFSI/DME is critical for achieving high ηCoulombic and dense, low-aspect-ratio Li electrode structures, (b) cell compression improves morphology and ηCoulombic likely by causing a denser and more self-limiting SEI film to form, (c) Li self-discharge occurs in 4M LiFSI/DME and this capacity loss mechanism can be suppressed with cell compression and by using LiAl0.3S protective coatings, (d) Li nucleation density improved in the presence of LiAl0.3S protective coatings even in the absence of cell compression, and (e) in-situ electrochemical characterization experiments which cannot integrate surface compression need to be designed carefully to ensure that the results are relevant to mechanisms occurring at compressed interfaces. RESULTS AND DISCUSSION Here we present real-time nanoscale visualization of Li deposition and stripping behavior through in-situ EC-STEM experiments in a recently developed12 and promising electrolyte, 4 M LiFSI/DME, with and without LiAl0.3S protective coatings. A cell designed and manufactured at Sandia National Laboratories’ Center for Integrated Nanotechnologies (CINT) was used in this work for the in-situ EC-STEM experiments, the CINT Electrochemical-Cell Discovery (CINTEC) platform.23 The CINT-EC design and previous applications have been described in detail elsewhere,18,

23, 24

but images are also provided in Figure 1 and details are discussed in the 6 ACS Paragon Plus Environment

Page 7 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

Experimental Section. The CINT-EC consists of top and bottom silicon chips with alignment beads centering the top and bottom SiN membrane windows. In these experiments, the ten electrodes were masked through an ALD Al2O3/sputtered SiO2 passivation layer (33 nm/7 nm, respectively), and smaller working electrodes (WEs) were lithographically patterned onto the chips. Larger electrodes were typically also patterned to serve as counter electrodes (CE) and pseudo-reference electrodes (p-RE); all experiments were performed with three electrodes. The CE and p-REs were typically composed of the same metal as the WE current collectors and they did not contain Li, as is typical for EC-STEM experiments. See more details in the Supporting Information. Beam Effects It is necessary to discuss the effects of the high energy electron beam used in transmission electron microscopy (TEM) and STEM experiments. The beam can potentially lead to electrolyte degradation and corresponding anomalous results, as has been discussed previously.18,

25

These effects are likely dependent on factors such as the beam operating

conditions, cell design, temperature, electrode composition, electrode design, and electrolyte composition. Previous studies using the CINT-EC platform have shown that the electron beam affected the Li morphology and electrochemical signatures in ethylene carbonate/dimethyl carbonate containing 1 M lithium hexafluorophosphate, even under relatively low electron flux conditions compared to standard TEM imaging (0.5 e-/Å2 per image, images every 15 s).18 To determine how the beam affected the Li morphology and electrochemistry specifically in 4M LiFSI/DME, we performed a series of in-situ EC-STEM experiments and present bright-field (BF) STEM imaging in Figure 2. As shown in Figure 2a, four Cu WEs were patterned onto the CINT-EC platform. Ten deposition and stripping cycles were run on each of three WEs and the 7 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 42

images shown were recorded upon conclusion of cycling for three different imaging conditions: (1) not imaged during Li cycling, Figure 2(b); (2) imaged only at the completion of each half cycle during open circuit conditions (5s per image), Figure 2(c); and (3) imaged every 15 seconds (5s per image) throughout cycling, Figure 2(d). The corresponding electrochemical traces are shown in Figure S1 along with further description of the experimental setup. During cycling, the WEs in Figure 2(b-d) were imaged with various accumulated electron fluxes ranging from 0 to ~20.16 e-/Å2. These images represent the imaging area, as well as the area just beyond the frame, to show beam damage that accumulated during raster of the probe (defined by a square shadow). It should be noted, that the liquid thickness within the CINT-EC platform was uniform over the window area, so additionally, added stress to the SiN membranes caused by Li deposits spanning between the two membranes could potentially account for the increased background signal about the electrode. Clearly, the contrast was reduced with increasing beam dose, Figure 2(c), and dark squares appeared for high doses where the STEM beam scanned the silicon-nitride windows during the electrochemistry, indicating carbon polymerization in 4M LiFSI/DME, Figure 2(d). In each of the following experiments, beam settings were carefully optimized to minimize the dose, as the influence of electrolyte degradation was not the focus of this work. Due to the beam sensitivity of the electrolyte, imaging in further experiments was performed only after each deposition and stripping half cycle (5 second exposure STEM images). It is likely that some beam damage occurred even within this limited exposure interval. Lithium Deposition and Stripping Behavior Many observations can be made through examining the STEM images in Figure 2 and the electrochemical results in Figure S1: (1) the Li deposits did not exhibit the uniform, dense 8 ACS Paragon Plus Environment

Page 9 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

morphology that had been previously observed under compression in coin cells with this electrolyte;12 (2) the ηCoulombic values were generally very low with very large standard deviations (average ηCoulombic is 29±29% for Figure S1(b), 17±16% for Figure S1(c), and 42±23% for Figure S1(d)) as expected based on the large amount of dead Li remaining after cycling; (3) dark particles decorated the Li deposits (Li is bright in BF-STEM imaging18,

26

); and (4) the Li

deposits composed of the dead Li that remained after 10 deposition/strip cycles, Figure 2(e), were largely consumed after 2 hours at open circuit conditions, as shown in Figure 2(f), indicating significant self-discharge in 4M LiFSI/DME at the nanoscale. The remainder of this work focuses on elucidating the causes of these unexpected observations. Morphology, ηCoulombic, and Compression A range of Li morphologies are evident in Figure 2 which do not resemble the morphology shown in coin cells with the 4M LiFSI/DME electrolyte under compression.12 The evolution of morphology in the EC-STEM experiment is more obvious in Figure S1 than in Figure 2, where images are presented after each half cycle for the experiment corresponding with Figure 2(c) on a Cu current collector. There are several factors that could contribute to the differences in morphology and ηCoulombic between these EC-STEM experiments and previous coin cell work.12 One important difference lies in the electrodeposited capacity. Li deposits were commonly observed at multiple points over the WE surface for the EC-STEM experiments, where the final capacity was low and only the early stages of growth were probed. In other experiments run to higher capacities (not shown), the Li grains began to overlap and a denser morphology was evident than what is presented in Figure 2. It is also possible that differences in the morphology and ηCoulombic shown in Qian et al.12 and our EC-STEM experiments arose due to differences in electrode size. The electrodes in the EC-STEM experiments have been previously 9 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 42

shown to behave as ultramicroelectrodes with limiting currents much higher than the currents used here.18 It has previously been suggested that the ultramicroelectrodes may be more likely to promote high-aspect-ratio dendrites and favor growth rather than nucleation. However, we should note that we typically observed several nucleation sites where new Li deposits grew in each cycle; Li deposited onto new current collector sites rather than the growth being limited to existing Li deposits, even on these small electrodes. While there are many factors unique to the EC-STEM environment which likely can affect the morphology and ηCoulombic, we hypothesized that differences between coin cells12 and the EC-STEM experiments could be attributed to differences in applied compression onto the deposited Li films in these two cell designs. To verify this assumption, we performed experiments in macroscale cells with and without interfacial compression (see Experimental Section for details). Cyclic voltammetry (CV) and SEM images (Figure 3) revealed significant differences in ηCoulombic and Li morphology with and without compression. CV measurements showed much higher ηCoulombic with compression than without, as shown in Figures 3(a) and 3(d), respectively. The Li morphology achieved with compression in Figure 3(b-c) was very similar to previous work in coin cells for galvanostatically-deposited Li;12 dense, uniform, and flat films were observed with compression. Conversely, mossy, high-aspect-ratio dendritic morphologies were observed without compression in Figure 3(e-f). Because the morphology was much denser and the ηCoulombic was much higher with compression, we conclude that compression is essential for achieving high ηCoulombic and favorable morphology in 4M LiFSI/DME. This conclusion is consistent with results from other electrolyte systems, in which Li deposition and ηCoulombic are dependent on cell compression.4, 19-22 The absence of compression in the EC-STEM experiments explains why the deposits formed with a different morphology and why they cycled at lower 10 ACS Paragon Plus Environment

Page 11 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

ηCoulombic than what has previously been observed in compressed, macroscale cells such as those in Figure 3(a-c) and in previous coin cell12 experiments. Although differences in electrode size and capacity may have accounted for some differences in the Li morphology between the macroscale (Figure 3) and microscale (Figures 2 and S1) current collectors, it is clear from Figure 3 that high ηCoulombic and a dense, regular morphology can only be achieved in 4M LiFSI/DME with compression. Previous literature has suggested that compression can improve Li deposition/stripping morphology, ηCoulombic, and cycleability.4,

22, 27

However, the effects of cell compression vary

greatly with electrolyte; cell compression has been shown to improve ηCoulombic greatly in some electrolytes but insignificantly in others. In particular, the salt composition has been shown to play a major role.22 Furthermore, some electrolyte additives improve ηCoulombic only with compression (vinylene carbonate, VC), causing cell failure after only a few cycles without compression.21 Conversely, another additive cetytrimethylammonium chloride (CTAC) has shown superior ηCoulombic only without compression;27 while compression greatly improved ηCoulombic in the electrolyte without CTAC, compression did not significantly improve ηCoulombic with CTAC. In summary, the relationship between ηCoulombic and compression is very complex and varies greatly with the electrolyte composition. The literature has not provided much discussion regarding the mechanism by which compression improves morphology, ηCoulombic, and cyclability. The observation that compression generally leads to denser Li deposits in many electrolytes is logical due to the constrained space that compression provides.4,

20, 22

Hirai et al. further suggested that compression improved

ηCoulombic by increasing the probability of reconnecting dead Li in low density morphologies.22

11 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 42

They also suggested that CTAC formed a surface film which inhibited reattachment of the dead Li, thus preventing compression from improving ηCoulombic in the presence of the additive.27 Ota et al. suggested that VC “effectively increased the cycling efficiency by pressurizing the electrodes.”21 However, denser morphology and reconnection of dead Li are not the only factors because the relationship between ηCoulombic and compression is complex and varies greatly with electrolyte composition, as discussed above. We propose that SEI formation (which is very dependent on the electrolyte) plays a major role in determining the relationship between compression and ηCoulombic. Evidence for this mechanism is provided by Figure 3. The reduction peak observed in Figure 3(d) around 1.75 V was only observed without compression in the first cycle. This experiment was repeated six times and this peak was evident on both Cu and Pt electrodes at both 1M and 4M LiFSI concentrations. Although SEI in most electrolytes forms at much lower potential, we assigned the peak at 1.75 V to SEI formation. The literature provides ample evidence that SEI formation occurs at high potential in LiFSI/DME (1.4-2.2 V) and that the high-potential SEI formation is due to concentration-dependent FSI anion decomposition.28-30 Note that running the experiment in Figure 3(d) with the LiTFSI rather than LiFSI salt resulted in an SEI formation peak at lower, more typical potentials (~0.5 V). Figure 3 shows that without compression, SEI formed readily and copiously (peak at 1.75 V), the Li morphology was not dense, and ηCoulombic was low. Conversely, with compression, the SEI peak was not visible, the Li morphology was dense, and ηCoulombic was much higher. Thus, SEI formed differently in compressed and uncompressed cells even on the first cycle in the absence of Li at potentials far above Li deposition. This suggests that cell compression changes how the SEI forms, likely leading to a denser and more self-limiting SEI that controls the 12 ACS Paragon Plus Environment

Page 13 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

morphology and serves to more effectively protect the electrolyte from contact with the electrode. The changes in SEI formation could arise from physical pressure on the SEI film itself or from mass-transport limitations in a compressed cell with a separator. The concept of a low-density SEI in the absence of compression is also supported by former EC-STEM work, where the authors observe a much thicker SEI than what has typically been described in the literature for macroscale cells;31 this discrepancy may be due to use of an uncompressed cell in those experiments. We propose that compression does more than just physically restrain Li deposits and allow reconnection of dead Li; compression also may change the microstructure of the SEI and promotes its ability to serve as an intrinsic protection membrane. This proposed mechanism explains why electrolyte composition (solvent, salt, and additives) affects the relationship between ηCoulombic and compression so significantly, given that SEI is highly dependent on the electrolyte composition. Understanding the mechanism by which compression controls Li morphology and ηCoulombic is critical for several reasons.

First, understanding how electrolyte performance

depends on compression offers the potential to optimize and further improve Li cycling performance.

Second, while coin cells and pouch cells can be compressed normal to the

electrode surfaces using external pressure, spirally-wound configurations (such as standard 18650 cells) cannot easily be evenly compressed.32 Third, because the cathode in batteries only expands by a few percent with Li intercalation but the Li metal anode expands and contracts drastically, compression can never be uniform and consistent in a commercial cell with an intercalation cathode. Typical half-cell experiments where Li plates and strips on each side provides relatively even expansion and contraction, so compression in half-cells is not representative of full cells with intercalation cathodes. Instead, compression force varies greatly 13 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 42

during cycling in practical cells with a Li anode and an intercalation cathode. Finally, several compression environments exist in a practical battery with a porous, ionically conducting separator. At the microscale, contact between the separator and the electrode surface varies even under high stack pressure due to separator porosity. The uncompressed microscale electrodes in the EC-STEM experiments are representative of the porous regions in the separator pressed against a macroscale electrode, as would be present in any commercial battery. Thus, understanding the dependence of a given electrolyte on compression is important for understanding and overcoming practical cell design and manufacturing complications. Decoration of Deposits with Dark Particles The next nanoscale observation based on Figure 2 is the presence of small dark deposits which decorate the Li grains, which are especially evident in Figure 2(e-f). Li appears bright in BF STEM imaging because it is the least dense material in the cell (less dense than the surrounding electrolyte).18, 26 The small, dark particles decorating the Li deposits are Cu metal that stripped from the Cu CE and deposited on the WE due to the absence of a Li-containing CE in these EC-STEM experiments. The absence of a Li-containing CE is discussed in greater detail in the Supporting Information and is typical in EC-(S)TEM experiments.26,

33, 34

Because Cu

deposited on the WEs during cycling, we also patterned Ni current collectors instead of Cu on the CINT-EC platform, as Ni is also an acceptable current collector material for Li metal anodes. Figure 4 shows images of Li deposits after various deposition and stripping cycles on Ni WEs. Similar to the Cu results shown in Figure 2, the Li morphology on Ni was irregular and the ηCoulombic was low (average ηCoulombic for 10 cycles was 18 ± 9%). Note that Figure S1 includes images after selected deposition and stripping cycles on the Cu current collectors described in

14 ACS Paragon Plus Environment

Page 15 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

Figure 2, which more directly compares to the macroscale cell results on Cu current collectors in Figure 3. Magnified images of the WE and CE shown in Figure 4 are presented in Figure 5. Closer inspection of the final image immediately after ten cycles, Figure 5(a), reveals dark particles around the Li deposits, similar to the EC-STEM results using Cu current collectors, Figure 2. Images of the Ni CE before and after cycling, Figure 5(c-d), reveal that the CE indeed was consumed during cycling and appeared to deposit on the WE during Li deposition. This result demonstrates the importance of incorporating a Li-containing CE in the electrochemical TEM liquid cells (see Supporting Information for more discussion). Direct Visualization of Self-Discharge Finally, comparison of Figure 5(a) immediately after cycling and 5(b) two hours later after a rest with the WE disconnected from the potentiostat, reveals that the bright white Li deposits were consumed during rest. In addition, Li grains were observed to gradually lose contrast over continuous deposition and stripping cycles. This was attributed to several possibilities: reconnection of dead Li which subsequently can strip, increased background contrast due to SEI formation, increased background contrast due to carbon polymerization of the electrolyte by the electron beam, increased background contrast due to increased cell thickness from large Li deposits pressing against the upper and lower SiN windows, or selfdischarge due to attack of the electrolyte on the highly reactive Li. These results are similar to those shown on Cu current collectors in Figure 2(e-f). The consumption of Li deposits during rest indicates that self-discharge occurs readily in 4M LiFSI/DME. The observed self-discharge behavior is the focus of the remainder of this manuscript. Suppression of Self-Discharge in 4M LiFSI/DME 15 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 42

Self-discharge occurs readily and quickly in 4M LiFSI/DME in a cell without compression, as evidenced in the EC-STEM experiments. The compact SEI that formed under compression in this electrolyte was shown to be helpful for minimizing ηCoulombic losses due to minimal SEI formation;12 however, perhaps such a compact SEI could not adequately prevent the Li deposits from interacting with the electrolyte and causing self-discharge during rest in the ECSTEM experiments. Self-discharge is a common problem in liquid electrolytes and results from corrosion reactions between the Li metal and the electrolyte due to inadequate SEI protection.35 Self-discharge has been previously shown to vary significantly with electrolyte composition such that electrolytes which cycle at higher ηCoulombic during cycling without rest often suffer less from self-discharge during rest periods.35 Therefore, because 4M LiFSI/DME cycles with such high ηCoulombic in the absence of rest periods,12 it is counter-intuitive that it readily suffers from selfdischarge. Although protective coatings are typically applied as artificial SEIs to improve ηCoulombic and morphology during cycling, we hypothesized that protective coatings are also likely to reduce self-discharge in rest periods during battery cycling. A previous publication has shown that ALD Al2O3 coatings can delay corrosion of pristine Li upon exposure to various environments before cycling.15 In the same work, the authors cycle Li-S coin cells with and without Al2O3 coatings. As the authors point out, because their Li-S cells include an excess of Li, their cycling experiments do not directly probe the effects of protective coatings on Li; rather, their ηCoulombic measurements reflect the electrochemical activity at the cathode, which was the metric of interest in their experiments. Therefore, to complement their work by isolating the effects of protective coatings on Li metal and to provide direct visualization of potential

16 ACS Paragon Plus Environment

Page 17 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

corrosion processes, we deposited 50 nm thick LiAl0.3S films by ALD on the CINT-EC platform and performed EC-STEM experiments. Figure 6 shows STEM images after select deposition and stripping cycles of Li on Ni electrodes with a LiAl0.3S protective coating. Li deposits formed in various morphologies, similar to the experiments without LiAl0.3S. Also, Ni appeared to redistribute underneath the LiAl0.3S (see red oval region). Li deposited below the LiAl0.3S protective coating, as evidenced by cracking in the Ni electrode in the location of Li deposits (see yellow ovals). This observation implies that the LiAl0.3S protective coating remained majorly intact at the Li-electrolyte interface during cycling and therefore continued to participate as a protective coating during cycling. The electrochemical response shown in Figure 6 differed from that characteristic of Li deposition and stripping (compare the shapes of charge-discharge curves without LiAl0.3S in Figure 4 and Figure S2, to those with LiAl0.3S in Figure 6 and Figure S2). In addition to the Li deposition and stripping plateau near -3 V versus the Ni p-RE, an additional reversible plateau is obvious at ~1.5 V higher than the Li+/Li0 couple. To ensure that the high voltage plateau was not an artifact of the EC-STEM experiments, we performed macroscale electrochemistry with and without LiAl0.3S coatings in beaker experiments versus a Li RE without a separator or cell compression. Figure S3 confirms a plateau ~1.5 V versus the Li+/Li0 couple only for experiments with the LiAl0.3S protective coating. Therefore, the observed plateau at ~1.5 V higher than the Li+/Li0 is reproducible and not an artifact of the EC-STEM experiment. Also interesting is that deposition was much more uniform with the LiAl0.3S coating, Figure S3(d), than without the coating, Figure S3(c), in these macroscale experiments indicating that Li nucleation site density was significantly increased at the original LiAl0.3S:metal (Ni) interface.

17 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 42

After the first two cycles, the higher voltage plateau shown in Figure 6(i-j) exhibits reversible electrochemistry as indicated in Table 1, whereas the Li+/Li0 plateau exhibits low ηCoulombic similar to experiments without the LiAl0.3S coating (Figure 4). The lower ηCoulombic in the first two cycles likely resulted from either SEI formation, which occurs at a similar potential, or irreversible Li insertion into the LiAl0.3S. Regardless, the ηCoulombic of the higher voltage plateau stabilized after the first two cycles, indicating a reversible Li transport mechanism through the coating was present in the remaining cycles. Note that the CE was consumed during this experiment, which led to high polarization at the CE during the final few cycles; thus the capacity was limited to only the higher voltage plateau for the 9th and 10th cycles and the Li+/Li0 plateau was only partially accessed during the 8th cycle. The reversibility of the higher voltage plateau was especially obvious in the final two cycles that did not access the Li+/Li0 couple, but Table 1 shows this was a general trend for all cycles. Because the higher potential plateau was so reversible, it is unlikely to be related to SEI formation (except for potentially during the first two cycles). Rather, the reversibility and the fact that it is only present with LiAl0.3S suggests that it may represent electrochemical lithiation (to saturation) and delithiation of the protective coating, especially since it is known that LiAl0.3S can be prepared with varied Li content. Because the ηCoulombic was not greater than 100%, it is unlikely that the coatings were delithiated beyond the initial, as-synthesized Li content. Furthermore, the reversibility of the LiAl0.3S coating and the consistent capacity (see Figure 6) related to the 1.5 V plateau indicates the continued participation with cycling of the LiAl0.3S in directing and stabilizing Li deposition. Table 1. Approximate ηCoulombic corresponding to the two plateaus shown in Figure 6(i-j).

Cycle Number

ηCoulombic ηCoulombic High Volt Li+/Li0 Plateau Plateau

ηCoulombic Overall 18

ACS Paragon Plus Environment

Page 19 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

1 2 3 4 5 6 7 8 9 10 Ave. ± σ

80 % 89 % 97 % 97 % 100 % 98 % 96 % 99 % 97 % 96% 95±6 %

60 % 47 % 44 % 50 % 29 % 8% 26 % 0% N/A N/A 33±21%

70 % 66 % 67 % 70 % 59 % 46 % 58 % 99 % 97 % 96 % 73±18 %

Examination of the EC-STEM images with LiAl0.3S protective coatings shows that selfdischarge was suppressed by the LiAl0.3S membrane. This is evident in the EC-STEM image presented immediately after cycling, Figure 7(a), compared to an image after four hours of rest when the WE was disconnected from the potentiostat and not exposed to the electron beam, Figure 7(b). In sharp contrast to the images shown in Figures 2, 4, and 5 without LiAl0.3S, Figure 7 shows no change in the morphology or amount of Li present after rest despite doubling the rest period in the LiAl0.3S experiment. The self-discharge shown in Figures 2, 4, and 5 suggests that the SEI formed during cycling was not sufficient to block corrosion reactions between the Li deposits and the electrolyte. Applying an artificial SEI such as the LiAl0.3S coating appeared to suppress this effect, which promoted greater Li protection than the intrinsic SEI formed in 4M LiFSI/DME. Because there are many differences between EC-STEM cells and macroscale cells, it is possible that the self-discharge results presented in Figures 2, 4, and 5 are an artifact of the ECSTEM experimental setup or conditions. The differences between our EC-STEM cell and coin cells include: (1) use of a non-Li-containing CE which could lead to byproducts in the cell that are not present in a more conventional cell format; (2) Ni or Cu metal redistribution in the cell

19 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 42

that decorates the Li deposits and could act as the more noble half of a galvanic couple, promoting corrosion of the Li; (3) absence of cell compression in the EC-STEM experiments, which we have shown decreased ηCoulombic and promoted non-uniform Li morphology; (4) Li morphology or SEI formation that is influenced by the electron beam, creating a more corrosionprone interface; and (5) use of ultramicroelectrodes in a constrained electrolyte volume leading to greater likelihood of reaction with byproducts. To determine whether the EC-STEM results are relevant in a more conventional cell format, coin cells were constructed with and without LiAl0.3S. These coin cells were fabricated with Ni as the WE and Li as the CE/RE so that the measurement of ηCoulombic was directly related to the Li deposition and stripping rather than electrochemistry at a cathode.

Four coin cells with LiAl0.3S and four without LiAl0.3S were cycled for 100 cycles at ±0.5 mA/cm2 with no rest between Li deposition and stripping for the purpose of verifying selfdischarge prevention with the protective coating. After every ten cycles, Li was deposited and subjected to a four-hour rest period at open circuit, after which the Li was stripped. Figure 8 shows that the ηCoulombic was low on the first cycle and then gradually rose to about 99% with cycling. This behavior is consistent with the previous report on 4M LiFSI/DME,12 though it differs slightly due to the fact that rest steps are interspersed in our cycling profile. ηCoulombic was generally similar between the cells with and without LiAl0.3S in the absence of rest between deposition and stripping. However, ηCoulombic drops significantly (below 96%) in the cells without LiAl0.3S for cycles with a four-hour rest between deposition and stripping. This Li inventory lost at open circuit was suppressed in cells with LiAl0.3S protection coatings, in agreement with the EC-STEM results. 20 ACS Paragon Plus Environment

Page 21 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

Despite the inconsistencies between EC-STEM experiments and coin cell experiments, the EC-STEM experiments were useful in identifying a self-discharge phenomenon relevant to coin cells. However, there are some noteworthy differences between the coin cell and EC-STEM results. First, self-discharge was more pronounced in the EC-STEM experiments than in the coin cell experiments, which may result from lack of compression. The SEI forms differently with and without compression (Figure 3), and it is likely a denser, more corrosion-resistance SEI forms under compression. Other differences include electrode size and capacity, higher byproduct sensitivity with ultramicroelectrodes, higher surface to volume ratio of Li deposits, Ni particles in the SEI, and byproducts from a non-Li-containing CE in the EC-STEM experiments. Second, cycles immediately following rest cycles (blue data points) in the coin cell data generally show higher ηCoulombic than before the rest cycle; meaning the ηCoulombic lost during the rest step was partially recovered. This recovery suggests that the self-discharge mechanism in coin cells includes both Li corrosion and disconnected Li that can reconnect upon further cycling. The reconnection was rarely observed in the EC-STEM experiments, where the deposited Li was generally completely consumed during rest and sometimes during subsequent deposition cycles. Note that dead Li alone cannot explain the loss in ηCoulombic with rest because there is no mechanism for dead Li to be generated during rest, unless corrosion reactions are also involved. Dead Li can either form during cycling when morphological changes are responsible for displacing Li or it can arise during rest if corrosion reactions lead to loss of connection to the electrode. Therefore, any losses in ηCoulombic during rest must have included corrosion, even if dead Li was also involved. Self-discharge was significantly suppressed in the coin cells with LiAl0.3S coatings, but there was still a measurable drop in ηCoulombic after rest. Contrastingly, the EC-STEM 21 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 22 of 42

experiments with LiAl0.3S coatings show no visual evidence of self-discharge after rest. The cycling capacity was significantly higher in the coin cell experiments than in the EC-STEM experiments which likely results in higher mechanical stress and pulverization of the LiAl0.3S coating during Li deposition and stripping in the coin cell experiments. More optimization must likely be performed to improve these coatings to be robust enough to withstand long-term cycling at capacities that are relevant for electrical energy storage applications. Regardless of the differences between EC-STEM and coin cell results, the self-discharge observations made in the EC-STEM cell were helpful for identifying and proposing a solution to the problem of Li selfdischarge in 4M LiFSI/DME. It is important to point out that to the in-situ characterization community that compression is an important factor in Li plating and stripping behavior, so great thought must be included in planning experiments that are valuable and relevant to the battery community given this limitation. The presence of a SiN window in close proximity to the WE may provide some compressive force when the deposits span the gap between the electrode and the top window, but they are likely to be more complaint than the separator in a macroscale cell. However, these microscale experiments are also useful as a representation of an area on the electrode that experiences incomplete contact with the separator, residing below a pore, at a macroscale battery interface. Furthermore, compression cannot be consistently applied in all battery cell designs and cannot be consistently applied during cycling in a cell with a Li metal anode and an intercalation cathode due to volume expansion mismatch. Therefore, experiments lacking compression can still be relevant to macroscale compressed cells in some configurations. Our results show that despite current limitations in EC-STEM cell design, these experiments can identify nanoscale mechanisms relevant to macroscale systems that are difficult to detect. CONCLUSION

22 ACS Paragon Plus Environment

Page 23 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

Li deposition and stripping behavior was studied through in-situ EC-STEM experiments and macroscale-cell electrochemistry in 4M LiFSI/DME. This work shows that: (a) mechanical cell compression is critical for achieving high ηCoulombic and dense, low-aspect-ratio deposited Li films using 4M LiFSI/DME; (b) in addition to reconnection of dead Li, compression affects how the SEI forms in early cycles, which in turn controls Li deposition and stripping performance; (c) self-discharge occurs in 4M LiFSI/DME and is exacerbated in the absence of cell compression; (d) self-discharge can be suppressed in 4M LiFSI/DME with LiAl0.3S protective coatings; (e) LiAl0.3S protective coatings increase the nucleation density of Li during deposition, which likely is responsible for improved Li morphology and ηCoulombic with coatings; and (f) despite several notable differences between macroscale electrochemical measurements and the EC-STEM experiments, EC-STEM experiments can identify mechanisms relevant to macroscale battery systems, which have complex compression environments. Although protective coatings are typically applied in an attempt to improve morphology, ηCoulombic, and SEI characteristics, this work reinforces that protective coatings may also prove helpful in suppressing self-discharge during realistic cycling in which rest is a prevalent part of the commercial use charge-discharge profile. EXPERIMENTAL SECTION Electrochemical STEM Liquid Cell. Several commercial EC-STEM liquid-cell holders have been developed and used to study battery applications.25, 26, 34, 36-40 However, we used a custom cell, designed and manufactured at Sandia National Laboratories and available for collaborative use at the Center for Integrated Nanotechnologies (CINT), for the in-situ EC-STEM experiments (CINT Electrochemical-Cell (CINT-EC) platform).23 Briefly, the cell consists of two chips: (1) a 3x5 mm Si bottom chip equipped with 8-10 electrodes in various patterns and a 30 µm circular 23 ACS Paragon Plus Environment

ACS Nano

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 24 of 42

diameter electron transparent silicon-nitride window for TEM viewing and (2) a 2x3 mm top chip equipped with an identical silicon-nitride window to the base and ports used to fill the cell with electrolyte (Figure 1a). Ten tungsten electrodes are located on the bottom chip’s siliconnitride window and are connected to bond pads by buried, insulated, and highly-doped polysilicon leads. Ruby crystals are used to align the silicon-nitride windows of the top and bottom chips through corresponding depressions. A raised barrier seal-ring surrounding the siliconnitride windows and fluid fill ports provides a reservoir for the electrolyte and defines the height between the SiN membrane windows during TEM imaging (~150 nm). To prepare the liquid cells for EC-STEM experiments, several processing steps were performed in CINT’s clean room. The bottom chip was masked by conformal coatings of ALD Al2O3 (~33 nm) then sputtered SiO2 (~7 nm) to passivate the tungsten electrodes. Electron-beam lithography was used to pattern WEs, CEs, and p-REs. The passivation layer was then HF-etched completely in the areas of the experimental electrodes and metal was evaporated on the patterned areas to act as current collectors for the WEs, as shown in Figure 1(b). The resulting chips were electrically insulated everywhere except in the locations where lithographically-patterned electrodes were deposited and on the bond pads. This processing was performed so that electrochemistry was isolated to only the small areas visualized during the experiment; thus the electrochemistry was directly correlated with the small electrodes, which were imaged wholly during the experiments. More details regarding the chip processing are provided in previous publications.18, 23, 24 The top chips were then aligned and epoxied to the processed bottom chips (Loctite Hysol 1C-LV), as shown in Figure 1(c). The epoxy was cured in an oven in air at 60-80oC overnight. The chips were wire bonded to chip carriers that were designed to be used in a 16-lead 24 ACS Paragon Plus Environment

Page 25 of 42

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Nano

custom TEM holder (Nanofactory Instruments AB). Once wire-bonded, the assembled liquid cells were transferred to a helium-filled glove box (MBraun). The chips were filled with 4M LiFSI/DME (