Molecular Layer Deposition of Crosslinked Polymeric Lithicone for Superior Lithium Metal Anodes

In this work, we for the first time developed a novel lithium-containing crosslinked polymeric material, a lithicone that enables exceptional protection effects over lithium (Li) metal anodes. This new lithicone was synthesized via an accurately controllable molecular layer deposition (MLD) process, in which lithium tert-butoxide (LTB) and glycerol (GL) were used as precursors. The resultant LiGL lithicone was analyzed using a suite of characterizations. Furthermore, we found that the LiGL thichicone could serve as an exceptional polymeric protection film over Li metal anodes. Our experimental data revealed that the Li electrodes coated by this LiGL lithicone can achieve a superior cycling stability, accounting for an extremely long cyclability of > 13,600 Li-stripping/plating cycles and having no failures so far in Li/Li symmetric cells at a current density of 5 mA/cm 2 and an areal capacity of 1 mAh/cm 2 . We found that, with a sufficient protection by this LiGL coating, Li electrodes could realize long-term stable cyclability with little formation of Li dendrites and solid electrolyte interphase. This novel LiGL represents a facile but effective solution to the existing issues of Li anodes and potentially paves a technically feasible route for lithium metal batteries.


INTRODUCTION
Lithium (Li) metal is among the most attractive anodes of rechargeable batteries, ascribed to its extremely high capacity of 3860 mAh/g and the lowest negative electrochemical potential (-3.04 V versus the standard hydrogen electrode). 1 The first practice of Li anode has been witnessed 40 years ago. 2 Due to its dendritic growth and continuous formation of solid electrolyte interphase (SEI), however, Li metal has been prohibited from commercialization. Along with an everincreasing interest in high-energy lithium metal batteries (LMBs) such as lithium-sulfur (Li-S) and lithium-oxygen (Li-O 2 ) batteries in recent years, stable Li metal anodes now are undergoing intensive investigation. A variety of technical strategies have been reported to date, such as three-dimensional (3D) Li-hosting frameworks, 3 electrolyte additives, 4 solid-state electrolytes, 5 and surface coatings. [6][7] Among these efforts, surface coating remains as an facile and effective route and there have been different inorganic and polymeric coatings developed via wet chemistry and vapor-phase processes. [6][7][8][9][10][11] In this regard, molecular layer deposition (MLD) recently has emerged as a new research thrust, featuring its accurately controllable growth of polymeric films with unparalleled conformality and uniformity at moderate temperatures of ≤ 300 o C. 6,[12][13] The first practice of MLD on Li anodes was exposed in 2018 and so far only a few polymeric MLD coatings have been reported, including polyurea, 14 AlEG (EG = ethylene glycol), [15][16] and ZrEG. 17 These polymeric films generally are electrically insulating and have better flexibility over inorganic films, contributing to better protection effects and thereby better performance of Li metal anodes. Stimulated by these compelling polymeric films via MLD, in this study we attempted to develop three new processes of lithicones through coupling lithium tert-butoxide (LTB, LiO t Bu) with one of three organic precursors, glycerol (GL), EG, and hydroquinone (HQ). Similar to other metalcones, 12 lithicones are polymeric lithium alkoxides with carbon-containing backbones, i.e., -Li-O-R-O-Li-. Among these three resultant lithicones (LiGL, LiEG, and LiHQ), we particularly investigated the protective effects of LiGL on Li metal anodes as well as its MLD growth. Very encouragingly, the LiGL films, compared to the other MLD coatings [14][15][16][17] reported to date, enable a much longer cyclability in Li/Li cells, accounting for over 13,600 Li-stripping/plating cycles (> 5,440 hrs) at an areal current density of 5 mA/cm 2 and around 5,500 Li-stripping/plating cycles (~5,500 hrs) at an areal current density of 2 mA/cm 2 both at an areal capacity of 1 mAh/cm 2 without any signs to failure so far. Compared to previous surface-modified Li anodes as summarized in literature, 8 to the best of our knowledge, these LiGL-coated Li electrodes in this work have exhibited the best performance to date. In Figure   S1 in Supporting Information, the cyclability of our LiGL-coated Li electrodes is compared to those Li electrodes coated by other MLD films and by other coatings reported in literature. Our LiGL-coated Li electrodes showed a remarkable improvement in cyclability.

2.1.
MLD Processes: Using a commercial MLD system (Savannah 200, Ultratech Inc., MA) with argon (Ar) as the carrier gas, three MLD processes of lithicones have been investigated in this study, i.e., LiGL, LiEG, and LiHQ. LTB (Sigma-Aldrich, USA) was preheated in a bubbler at 150 o C for a sufficient vapor supply, while EG (Sigma-Aldrich, USA), GL (Sigma-Aldrich, USA), and HQ (Sigma-Aldrich, USA) were maintained in stainless steel cylinders at 40, 150, and 150 o C, respectively. This MLD system was integrated with an Ar-filled glove box (having an oxygen and water concentration lower than 1 ppm) and this integration has greatly expanded the fabrication capability of the MLD system to many air-sensitive materials. In these three processes, LTB as the Li source was commonly used to couple with GL, EG (Sigma-Aldrich, USA), and HQ at 150 o C, respectively. The timing sequence of a single MLD cycle was typically in a sequence of t l -t 2 -t 3 -t 4 , corresponding to the LTB dose, the first Ar purge, the GL/EG/HQ dose, and the second Ar purge, respectively. The precursor dosing time t 1 was optimized as 3 s, while t 3 was optimized as 0.15 s for EG, 2 s for GL, and 2 s for HQ. The purge time of t 2 and t 4 was optimized as 60 s for all these lithicones. During the MLD processes, the Ar gas flow was remained at 20 sccm.
These three MLD processes have been monitored using an in situ quartz crystal microbalance (QCM). The QCM studies were conducted using a gold sensor crystal (Inficon, USA). The crystal was sealed at the center of the MLD reactor lid. The doses of the different MLD precursors cause surface reactions with the growth of the targeted lithicones on the crystal surface. The growth of these three lithicones was monitored by the QCM to acquire timeresolved mass changes in ng/cm 2 and was recorded in digital data. To establish a uniform starting surface for these three MLD processes, we deposited an Al 2 O 3 layer on the QCM surface via atomic layer deposition (ALD, an analogous technique of MLD exclusively for inorganics 12 ) using trimethylaluminum (TMA) and water as precursors with the timing sequence 0.05-10-0.025-10 s.
We observed the deposition of LiGL films over N-GNS using a scanning electron microscopy (SEM, XL30, Philips FEI) to determine the film thickness and thereby the growth per cycle (GPC). The SEM is equipped with an energy-dispersive X-ray spectroscopy (EDX), which was used to detect the composition of the LiGL films. Synchrotron-based X-ray diffraction (XRD) was used to determine the crystallinity of the LiGL films and performed at the beamline 12-ID-D

The Growth of LiGL Films
In this study, we attempted three precursor couples for growing lithicones, i.e., LTB-GL, LTB-EG, and LTB-HQ. Figure 1a illustrates the growth principle of these lithicones, exemplified by the MLD LiGL process using LTB and GL as precursors. One LiGL MLD cycle consists of four repeatable steps: LTB dose/1 st purge/GL dose/2 nd purge. We controlled the MLD cycles to achieve a desirable film thickness. We measured the growth of these MLD processes using an in situ QCM. [24][25] To establish a repeatable uniform starting surface, we pre-deposited an Al 2 O 3 surface on the QCM crystals using ALD (inset of Figure 1b and 2b), and LiHQ ( Figure S2c and 2d). We were particularly interested in LiGL, for GL has three hydroxyl groups and is likely to form a crosslinked polymeric structure. 26 Our QCM measurements also revealed that the LiGL MLD process enables a much faster growth per cycle (GPC) (Figure 1b) than those of LiEG and LiHQ MLD processes. Consequently, we devoted to investigating the MLD process of LiGL and its effects on Li anodes. MLD QCM measurements in Figure 1b disclose that the LiGL growth on Al 2 O 3 is nearly linear but can be divided into two regions: initiation and stable growth regions. Figure 1c and 1d reveal more details in a single MLD cycle during the initiation and stable growth region, respectively, in which the doses of LTB and GL are signified by two different colored bars. Each dose of these two precursors caused some mass gain (m 1 or m 2 as shown in Figure 1d) on QCM. The average mass gain (m = m 1 + m 2 ) is ~200 ng/cm 2 /cycle in the initiation region (~30 cycles starting on an Al 2 O 3 film, Figure S3) while ~320 ng/cm 2 /cycle in the stable growth region ( Figure S3). The stable region exhibits a highly repeatable GPC. We deposited the LiGL films over one type of N-GNS ( Figure   2a) to determine its GPC at 150 o C. The N-GNS features its high surface area and thin wrinkles of < 3 nm. 27 Observing the thickness changes of the wrinkle of the N-GNS after 20 cycles ( Figure 2b) of the LiGL deposition using an SEM, we could conclude that the average GPC of the MLD LiGL is ~2.7 nm/cycle, which is among the highest GPCs of all the MLD processes reported to date. In addition to the SEM images, we conducted elemental mapping over the 20cycle LiGL-coated NGS using an EDX (Figure 2c), which shows the distributions of N, C, and O.
N is from N-GNS, C is from N-GNS and LiGL, and O is from LiGL. Consequently, EDX mapping revealed that the MLD LiGL coating over the N-GNS is very conformal and uniform.
According to our experience in MLD, we postulated the overall reaction of the MLD LiGL in Equation 1 as follows: Thus, the LiGL is supposed to have a unit structure of (CH 2 CHCH 2 )(OLi) 3  XPS spectra shows only one peak at 53.3 eV and this should be attributed to Li-O. 29,36 According to the XPS analyses, the deposited LiGL contains 31.69 at.% of Li, 28.73 at.% of C, and 39.59 at.% of O (more details are in Table S1). The element contents of Li, C, and O are basically consistent to our postulation on the LiGL unit structure, (CH 2 CHCH 2 )(OLi) 3 . However, we also noticed that there are less C and more O than those of our postulation. This might be due to some side reactions and incomplete surface reactions. Such cases have been widely reported in previous ALD and MLD studies. 12,37 To further gain some insight on the evolution of the LiGL films with film thickness, we utilized XPS to conduct depth profiling over a 75-MLD-cycle LiGL film on an Si wafer ( Figure S4). It was revealed that the composition of the LiGL film varies with film thickness. For example, the Li content changes from ~20 at.% at the top film surface to ~50 at.% at the interface between the LiGL film and the Si wafer. Two reasons might have contributed to this phenomenon: (i) the influence of substrates (e.g., Si wafers) and (ii) the influence of the successor MLD cycles on the predecessor MLD cycles. The former has been widely noticed while the latter was also reported previously. 38 We will invest some special effort on these topics in our future research. Additionally, we did some preliminary study on the air stability of the as-deposited LiGL and measured the XPS spectra of an LiGL film exposed in air for 1 h (more details are in Table S2). The effects of 1-h air exposure were compared in Figure   S5 and discussed in Supporting Information. The information revealed that the as-deposited LiGL is sensitive to air to some extent, such as water. In addition, synchrotron-based XRD has been conducted on the LiGL films grown on N-GNS and revealed no pronounced crystallinity with the LiGL films (very broad diffuse scattering peak shown in Figure S6). Thus, the LiGL films are amorphous. Moreover, we computed the amorphous LiGL bulk structures using AIMD simulation, and the calculated radial distribution function (RDF) confirmed the dominance of Li-O, C-O, C-H, and C-C bonds in the LiGL films ( Figure S7).

The Superior Performance of the LiGL-Coated Li/Li Symmetric Cells
To To further verify the protective effects of the MLD LiGL coatings, we applied a much higher current density of 5 mA/cm 2 (Figure 3b). Our tests revealed that a bare Li/Li cell (i.e., LiGL-0) sustained a low overpotential of < 130 mV in the first ~350 cycles, but then exhibited a constant increase of cell overpotential and failed after ~1,300 Li-stripping/plating cycles. In comparison, the LiGL-15/LiGL-15 cell has remained a low overpotential of < 130 mV in the first 500 cycles, but then had a continuously evident increase in cell overpotential and failed after ~2,000 cycles.
The results indicated that both the LiGL-0/LiGL-0 and LiGL15/LiGL-15 cells probably suffered from uncontrollable formation of SEI or/and Li dendritic growth. In contrast, the LiGL-20/LiGL-20 cell sustained a stable overpotential of <130 mV in ~1,000 cycles and then exhibited a gradually increasing overpotential from ~130 mV at ~1,000 th cycle to ~370 mV after 5,000 cycles. More details have been shown in Figure S9. Very impressively, the LiGL-60/LiGL-60 cell has accomplished over 13,600 Li-stripping/plating cycles with a much smaller overpotential from ~130 mV at ≤1,000 cycles to ~220 mV after 13,600 cycles. Very encouragingly, the LiGL-60/LiGL60 cell has not shown any failure and is still under testing. The excellent performance of the LiGL-coated Li/Li cells was compared with the results reported in literature. 8 To the best of our knowledge, our LiGL/LiGL cells have set a record on the best cyclability reported so far. We even further investigated the effects of LiGL coatings at a much higher current density to 7.5 mA/cm 2 at an areal capacity of 1 mAh/cm 2 . Our results ( Figure S10) show that the LiGL-90/LiGL-90 cell performed the best. These results seem to suggest that a thicker coating is more favorable for achieving stable cyclability at a higher current density. As for the underlying mechanism for the better effects of thicker LiGL coatings, we will discuss in the following parts in this study.

Understanding the protective effects of LiGL coatings
To understand the protective effects of the LiGL coatings, we investigated some cycled cells: LiGL-coated and uncoated (bare). In Figure 4a Thus, we concluded that the thicker the LiGL coatings, the better the protection effects. Very excitingly, the cross section of the LiGL-60 is nearly intact without evident SEI layers and Li dendrites after cycling (Figure 4b).
Furthermore, we employed XPS depth profiling to investigate the composition evolution with film depth on three Li electrodes, one bare Li after 10 Li-stripping/plating cycles (Figure 4c (Figure 5a(ii)-(iv)). One can clearly see from Figure  5a(iv) that these micro-wells contain many dendritic structures (or micro-pillars). We postulated that the craters were the areas that have stripped Li first and then ceased while the bumps were the areas that did not strip Li at the very beginning but became the new areas for stripping after the craters ceased stripping. Thus, the stripping process on bare Li electrodes is not uniform and the stripping areas change with time. On the other hand, the opposite bare Li after a 24-h plating was also observed (Figure 5b). Very strikingly, there is a large amount of dendritic Li deposited on the bare Li surface (Figure 5b(i)). It could be seen from Figure 5b To further clarify the effect of the LiGL coating thickness, we also observed the bare Li/Li and LiGL-20/LiGL-20 cells after one stripping (or plating) at a current density 2 mA/cm 2 and a capacity of 1 mAh/cm 2 ( Figure S14). No surprise, the bare Li was covered by dendritic Li on the plating side ( Figure S14a) while the bare Li was decorated with craters and bumps on the stripping side ( Figure S14b). In contrast, the LiGL-20 electrodes were clean. Specifically, there were no dendritic structures on the plating side ( Figure S14c) and there had no crates and bumps on the stripping side ( Figure S14d). Compared to the LiGL-60 coating, however, the LiGL-20 coating was prone to break into smaller pieces compared to the LiGL-60 coating ( Figure 5 and 6).
Thus, thicker LiGL coatings are beneficial to protect Li metal electrodes and thereby realize better cyclability, as shown in Figure 3 and 4.
To achieve high energy LMBs, particularly, thin Li films (50 μm or thinner) are required. [39][40] In this regard, we have confirmed that our LiGL coatings could be scaled up at ease and have been deposited on large-scale thin Li films (~ 50 μm) uniformly for making pouch cells. The results from pouch cells will be reported somewhere else.

Understanding the properties of LiGL coatings
Evidently, the LiGL coatings have exceptional protection effects on Li metal electrodes.
Therefore, their properties are of significance, such as mechanical properties and conductivities.
In this regard, we employed AFM to determine the topography and nanomechanical properties of and (iii) shows no porosity, and there is a significant difference in the average Young's modulus (3.24 GPa) and adhesion force (77 nN) between this and the porous areas of the coating. We also investigated the LiGL-200 coating which showed much more uniform topography than the LiGL-100 coating (Figure 7b(i)). Its nanomechanical properties are also consistent throughout the mapped area. Figure 7c shows the comparison of nanomechanical properties of the LiGL-100 and LiGL-200 coatings. The LiGL-100 coating shows higher average Young's modulus and adhesion than those of the LiGL-200 coating, due to the porosity and substrate effect on the thinner coating.
Compared to the Young's modulus of Li metal (7.82 GPa), 41 these results revealed that the LiGL coatings are moderate in their mechanical properties. This underlies why the LiGL coatings are vulnerable to mechanical forces and easy to fracture. We conducted a comparative study through assembling a bare Li/Li cell and a LiGL-20/LiGL-20 cell and then opening them to observe their surface changes. We found that the bare Li after the press in assembling changed from a flat surface ( Figure S15a) to a waved surface ( Figure S15b and S15c) locally covered with numerous scaled patterns ( Figure S15d-S15f). These patterns are several microns in size. On the other hand, we observed that the LiGL-20 electrode after the press in assembling has broken from a smooth surface ( Figure S16a) into scaled patterns of several microns (Figure S16b-S16d). All these results clearly evidenced that the LiGL coatings were broken mainly during mechanical assembling but fairly stable during Li-stripping/stripping cycles. These also explained that thicker LiGL coatings could better protect Li.
In addition to the mechanical properties of the LiGL coatings, their electrical and ionic conductivity are of particularly significance. In this regard, we conducted experimental measurements and computational simulations. From DFT calculation, the amorphous LiGL is found to be electronically insulating with ~ 3.0 eV in band gap, as shown in the electronic density of states close to the Fermi level (Fig. 8a). To explore the stability of electronic properties of the amorphous LiGL bulk at room temperature, we investigated the evolution of the highest occupied molecular orbital/band (HOMO), Fermi level, and the lowest unoccupied orbital/band (LUMO) of the amorphous LiGL bulk based on the structure changes in AIMD simulation for 15 ps (Figure 8b). Our simulations revealed that, as shown in Figure 8b ( Figure S17). Interestingly, this insulating merit can be persisted even at higher temperatures (e.g., ~550 K) with a slightly smaller band gap ~ 2.2 eV ( Figure S18). Thus, we believe that the electronically insulating nature of the LiGL films must have helped suppress the chemically reactive metallic Li dendrites even at elevated temperatures. Our experimental measurements also verified the computational results and further revealed that the LiGL coating are ionically conductive. All these results will be reported in a following systematic study.
We further studied the evolution of cell impedance with cycles ( Figure 8c

The proposed mechanism of the LiGL protection effects on Li electrodes
Based on the afore-discussed experimental and simulation results, we proposed the following mechanisms of Li-stripping/plating for the bare Li/Li cell and the LiGL/LiGL cell, as illustrated in Figure 9a and 9b, respectively.
As shown in Figure 9a for the evolution of the cross section of a bare Li electrode, an Li chip is initially smooth (Figure 9a1) but becomes bumpy after the assembling press (Figure 9a2). At the same time, the Li chip surface has been formed with an SEI layer, due to its contact and reaction with the electrolyte (Figure 9a2). The surface bumps of the Li chip are prone to take the priority to start an Li-stripping earlier (Figure 9a3). With the depletion of the Li bumps, the surrounding areas become new bumps and take the priority to continue the stripping while the former bumps become craters covered a layer of SEI and some residuals of SEI from the depleted bumps Eventually, the cell may be dried and have a significant increase of cell impedance and cell overpotential.
Different from the bare Li chip as illustrated in Figure 9a, the LiGL MLD process can form a uniform coating over the Li chip surface (Figure 9b1). After the mechanical assembling press, the LiGL-coated Li chip becomes bumpy (Figure 9b2). During this press, the LiGL coating breaks into small pieces. The fractures contact the electrolyte and forms an SEI layer. In comparison, the majority of the Li chip surface is still covered by the LiGL coating and only a very small part of the surface is covered by an SEI layer. The LiGL coating is electronically insulating while ionically conductive. As a result, the stripping starts from the bumpy areas ( Figure 9b3) but the Li chip surface become even quickly with the depleted bumps ( Figure 9b4).
Owing to the uniformity and the exceptional properties of the LiGL coating, the following stripping is remarkably even over the whole Li chip. Due to the uniform properties of the LiGLcoated Li chip surface, the following plating is also even without any further SEI formation ( Figure 9b5). Consequently, the LiGL-coated Li chip can realize long-term stable cyclability without evident consumption of Li and electrolyte. Thus, the cell overpotential can sustain stable with extended cycles.

CONCLUSION
In summary, we for the first time developed a novel lithicone, LiGL, in this study, which can be deposited uniformly and conformally with an accurately controllable GPC at a moderate temperature of 150 o C. Significantly, this LiGL MLD has a decent average GPC of ~2.7 nm/cycle and shows exceptional protection effects on Li electrodes, i.e., remarkably suppressing Li dendrites and mitigating SEI formation. Furthermore, we have clearly explored the protective mechanism of the LiGL coatings experimentally and computationally. Our computational simulations and experiments revealed that the MLD LiGL films are electrically insulating and ionically conductive. In addition, experimental results revealed a moderate stiffness of the LiGL coating. All these properties of the LiGL coatings underlie their excellent protective effects on Li metal electrodes. This work represents a facile solution to achieve high-performance Li metal anodes.

Data Availability
All data needed to evaluate the conclusions in the paper are present in the paper and the Supplementary Materials. Additional data related to this paper may be requested from the corresponding authors.

Conflict of Interest
The authors declare no conflict of interest.          the Li-stripping/plating mechanism of LiGL-coated Li.