In situ observation of cracking and self-healing of solid electrolyte interphases during lithium deposition

The growth of lithium (Li) whiskers is detrimental to Li batteries. However, it remains a challenge to directly track Li whisker growth. Here we report in situ observations of electrochemically induced Li deposition under a CO 2 atmosphere inside an environmental transmission electron microscope. We ﬁnd that the morphology of individual Li deposits is strongly inﬂuenced by the competing processes of cracking and self-healing of the solid electrolyte interphase (SEI). When cracking overwhelms self-healing, the directional growth of Li whiskers predominates. In contrast, when self-healing dominates over cracking, the isotropic growth of round Li particles prevails. The Li deposition rate and SEI constituent can be tuned to control the Li morphologies. We reveal a new ‘‘weak-spot” mode of Li dendrite growth, which is attributed to the operation of the Bardeen-Herring growth mechanism in the whisker’s cross section. This work has implications for the control of Li dendrite growth in Li batteries.

Both the tip-growth and bottom-growth modes of Li dendrites have been reported in Refs.[15,16,[31][32][33][34][35][36].Dendritic metal deposition can involve the long-range diffusion-limited mechanism associated with a well-known characteristic ''Sand's time" [31][32][33].In this case, Li dendrites mainly form via the tip-growth mode caused by a severe scarce of Li ions in liquid electrolytes.When the supply of Li ions is adequate, the bottom-growth mode can become predominant.During this process, the highly reactive metallic Li usually causes quick formation of a stable SEI film on its surface to prevent further reaction with liquid electrolytes [15,16,[34][35][36][37].Such growth can be strongly influenced by cracking and selfhealing of the SEI film on whisker's surface [38].On the other hand, protrusion of a Li whisker from the surface of the Li metal anode is usually associated with cracking and self-healing of the SEI film on the surface of bulk Li metal [39,40].Despite these mechanistic insights into Li dendrite growth, the dynamic processes of SEI cracking and self-healing have not been clearly understood.Therefore, direct observations of real-time cracking and self-healing of the SEI film are critically needed, in order to understand the mechanism of Li dendrite growth and ultimately find an effective means to control Li dendrite growth.
In situ experiment is particularly useful for revealing the dynamic structural evolution during the growth of Li whiskers.For example, in situ observations of Li growth in a glass cell optical microscope [15,16,35], liquid cell transmission electron microscope (TEM) [36,39,41], or plastic cell scanning electron microscope (SEM) [34,42] have been reported to investigate the growth mechanism of Li deposits.However, the nanoscale structural changes, such as cracking and self-healing of the ultrathin SEI film, have not been reported.In situ studies of SEI evolution are extremely challenging because the SEI layer is very thin (usually less than 20 nm) and tend to be ''living layers; that is, they are always changing" [43].
Generally, the SEI film consists of a dense inner layer with the main constituents of inorganic Li 2 CO 3 , Li 2 O and LiF, and a porous outer layer with organic Li compounds such as ROCO 2 Li, ROLi, and others [17,18].As the inner layer is directly in contact with metallic Li, Li 2 CO 3 and Li 2 O are the important SEI components that can strongly influence the electro-chemo-mechanical responses of a growing Li deposit [44].In situ imaging of the dynamic formation, fracture, reformation of either a Li 2 CO 3 or Li 2 O film on the surface of a growing Li deposit could provide crucial understanding of dendritic Li growth under the confinement of surface SEI films in real LMBs.

Experimental
The nanobattery setup for Li electroplating is a two-probe system in a spherical aberration-corrected environmental transmission electron microscope (ETEM), as schematically shown in Fig. 1.A multiwall carbon nanotube (MWCNT) obtained via an arc-discharge method [45] was glued to an aluminum rod with silver epoxy and was used as the cathode for Li electroplating.Metallic Li on a W tip was used as the Li anode.The Li 2 CO 3 or Li 2 O formed on the surface of the metallic Li under CO 2 gas or O 2 gas, respectively, in the ETEM chamber served as the solid electrolyte to conduct Li ions.Both the MWCNT cathode and the Li anode were mounted onto a TEM-STM (Scanning Tunneling Microscopy) holder (Pico Femto FE-F20) inside an argon gas filled glove box.Then the holder was sealed in a custom-made air-tight bag filled with dry argon and transferred to the ETEM.Prior to the experiment, high-purity CO 2 (99.99%) or O 2 was introduced into the ETEM chamber with a pressure of 1.0 mbar (1 mbar = 100 Pa).The pressure of CO 2 or O 2 gas in the TEM column was maintained at 1 mbar throughout the operation.

Results and discussion
To observe the dynamic processes occurring in the SEI film of Li 2 CO 3 on metallic Li, an in situ electrochemical plating device was constructed inside an aberration-corrected ETEM.As shown in Fig. 1, a piece of Li metal on a W tip was used as the anode; a Li 2 CO 3 layer on the surface of Li metal (formed within a CO 2 atmosphere inside ETEM) served as the solid electrolyte; and arc discharged MWCNTs acted as the cathode.When an external voltage was applied to this electrochemical plating device, metallic Li was deposited onto the surface of MWCNTs.Meanwhile, CO 2 in the ETEM chamber reacted with freshly deposited Li to form a surface layer of Li 2 CO 3 , electron microscopy characterization of which has been reported previously [46].The formation of such a Li 2 CO 3 layer on the deposited Li is considered to represent similar processes of SEI formation on the surface of Li deposits in liquid organic electrolytes of practical LMBs or in a solid state Li-CO 2 battery, which involve electrochemical reactions between Li deposits and electrolytes or the CO 2 gas, respectively [47].As shown in Fig. 2a, b and Movie S1 (online), when a negative potential was applied to the MWCNT cathode, Li ions from the Li metal anode diffused through the solid electrolyte layer of Li 2 CO 3 on the surface of the Li metal anode to lithiate the MWCNT.As lithiation of the MWCNT became saturated, a Li sphere emerged at the contact point of MWCNT/Li 2 CO 3 /CO 2 via an electrochemical plating process of Li + + e À ?Li.Meanwhile, an SEI film, in the form of a Li 2 CO 3 shell, formed on the surface of the Li sphere.This Li 2 CO 3 shell exhibits darker contrast than the Li core and thus can be recognized during in situ TEM observation [46].It was formed via electrochemical/chemical reactions between Li and CO 2 gas.As Li plating proceeded, the Li sphere grew from a diameter of ~200 (Fig. 2b) to ~400 nm (Fig. 2d), and the thickness of the SEI shell increased from 5 to 9 nm correspondingly, although the increase of the SEI shell thickness slowed down with lapse of time which saturated at a thickness less than about 20 nm [48].Under a high magnitude of applied potential (~À1.0 V), the growing Li sphere triggered the formation of a local crack (marked by a yellow arrowhead in Fig. 2e) in the SEI shell.Subsequently, two symmetric cracks (marked by a pair of yellow arrowheads in Fig. 2f) appeared in the SEI surface layer near the bottom of the Li sphere, suggesting the formation of a three-dimensional (3D) ring crack in the SEI shell.As a result of SEI cracking, the local mechanical constraint from the SEI shell was released, so that the Li sphere began to elongate along the upright direction (indicated by a yellow arrow in Fig. 2f).Such upright growth (indicated by a yellow arrow in Fig. 2g) continued, resulting in a long and straight Li whisker with a uniform cross-sectional width (Fig. 2g-j).It is seen that the SEI film formed at the bottom of the Li whisker is markedly thinner than that on the top of this Li whisker (Fig. 2g-j).During the growth of the Li whisker, the top hemispherical head remained unchanged, indicating the major growth occurring near the bottom region of the Li whisker [36].Protected by the thin Li 2 CO 3 shell, the Li whisker was relatively stable [48], offering the opportunity to obtain an electron diffraction pattern (EDP) of the electron-beam sensitive Li whisker with little irradiation damage.The EDP of this newly formed Li whisker (Fig. 2l) is indexed as the [011] zone axis of the body-centered cubic (BCC) Li whisker (PDF No. 15-0401) with the h1 11i growth direction (Fig. 2k).The dark field image (Fig. 2m) confirms the single crystalline structure of this Li whisker.
The above Li growth process was repeatable with more examples shown in Figs.S1, S3-S5, Movies S2-S5 (online).Occasional kinking along single-crystal Li whiskers was observed (Figs.S1, S2a online).It is interesting to note that even after the kink formation, the whisker remains a single crystal with the same crystallographic direction of growth (Fig. S2 online), which is different from the kinked Li whisker formed in the liquid organic electrolyte [49].This indicates the preferred growth along a specific crystallographic orientation such as h1 1 1i in the BCC Li, considering that there are several equivalent directions within a family of crystallographic orientations.Time lapse monitoring of the whisker growth (Figs.S1, S3, S4, Movies S2-S4 online) shows that in the projected TEM image, the newly formed SEI film was often thinner on one side than on the opposite side before the formation of a kink (Figs.S1h, j, S3v, S4f online).This result suggests that the kink formation can be ascribed to non-uniform Li deposition in the whisker.
The above observations show that the growth of Li whiskers initiated from a 3D ring crack formed in the SEI shell.Such directional growth requires that Li growth along the axial direction of the Li whisker (~14 nm s À1 ) overwhelms self-healing of the ring crack, presumably with the aid of fast reactions between locally exposed Li and CO 2 in the ETEM chamber.We discovered that under a low magnitude of applied voltage (~À0.5 V), the crack in the SEI layer healed quickly, such that no whisker growth was observed, as exemplified in Fig. 3 and Movie S6 (online).In this case, cracks (marked by red, green, blue and yellow arrowheads) emerged and then healed quickly as the spherical Li particle grew.This growth mode contrasts with that in Fig. 2 where a stable ring crack formed at the bottom of the Li particle, leading to the directional growth of a Li whisker.Note that the healable cracks in Fig. 3 were mainly formed far from the Li source of the Li metal anode in the present electrochemical plating setup.As seen from the dynamic processes of cracking and self-healing in the SEI film (Fig. 3a-d, red arrowheads), cracks formed at the top of the Li particle, due to mechanical damage of the Li 2 CO 3 shell.Such damage stemmed from the buildup of large hoop tensile stresses in the Li 2 CO 3 shell during expansion of the growing Li particle.Crack formation in the Li 2 CO 3 shell exposed the freshly deposited Li to the CO 2 gas in the ETEM chamber.This led to quick local self-healing of the Li 2 -CO 3 shell through the electrochemical reduction of CO 2 via 4Li + 3CO 2 ?2Li 2 CO 3 + C or Li + 2CO 2 ?2Li 2 CO 3 + CO [50].Hence, the Li whisker was not formed when the magnitude of applied voltages was low, although two symmetric cracks in the TEM image (indicating a 3D ring crack) were temporarily present, as marked by blue arrowheads in Fig. 3j-l.Apparently, the slow Li deposition provided enough time for the locally-cracked SEI film to self-heal, thus suppressing Li whisker growth.Similar Li deposition processes were observed when applied voltages were low (Figs.S6-S8 and Movies S7-S9 online), demonstrating the repeatability of the growth of spherical Li particles under low driving forces.Moreover, the Li 2 CO 3 -coated Li particles (Fig. 4, Fig. S9, and Movies S10-S11 online) were obtained when the applied voltage for Li deposition was further reduced to À0.2 V.These results indicate that the growth of spherical Li particles can be effectively controlled by applied voltage.The thickness of the Li 2 CO 3 SEI shell was uniform and gradually increased as the Li particle grew (Fig. 4, and Fig. S9 online), indicating the continued reaction of Li with CO 2 to thicken the Li 2 CO 3 shell.
The above experiments demonstrate that at high applied voltages, Li deposition overwhelms crack self-healing, leading to the directional growth of Li whiskers; in contrast, at low applied voltages, crack self-healing is sufficiently fast, leading to the isotropic growth of Li particles.Based on these results, we explored the dynamic control of Li deposit morphologies by tuning applied voltages.An example of the dynamically tunable Li morphology is pro-vided in Fig. 5, and Movie S12 (online).In this case, a Li whisker grew initially under a high applied voltage of À3.0 V, and this Li deposit subsequently became a nearly sphere (marked by yellow dashed lines in Fig. 5b-h) after the applied voltage was lowered to À1.0 V.This result indicates that when self-healing is sufficiently fast, the growth of Li particles predominates.It is also noted that the aspect ratio of another slowly-growing Li whisker (tilted to the left) became smaller because of a low voltage applied (Fig. 5c-h).In addition, we conducted a cycling experiment and measured the reversibility of the electrochemically grown whiskers.As shown in Fig. S10 (online), when a high negative potential was applied to the MWCNT cathode, a lithium sphere emerged, which evolved into a whisker after 1785 s.When the voltage was reversed, the whisker shrank continuously until a Li 2 CO 3 shell was left (Fig. S10l online).Upon applying a negative bias on the CNT again, the lithium plated back into the Li 2 CO 3 shell again (Fig. S10r online), and the whisker shrank again at a reversed bias (Fig. S10x online).The discharge/charge processes were repeatable for two cycles (Fig. S10a-x).
We also investigated the impact of SEI composition on the morphology of Li deposits.When the SEI film was composed of Li 2 O rather than Li 2 CO 3 , we observed the favorable growth of Li particles rather than Li whiskers.For example, we conducted Li deposition experiments in an O 2 atmosphere, producing the surface SEI film of Li 2 O instead of Li 2 CO 3 .As shown in Fig. 6, Fig. S11 and Movies S13, S14 (online), only spherical Li particles formed even when the applied voltage was high (À1.5 V), which is different from the voltage-dependent formation of Li whiskers and particles in a CO 2 atmosphere.The EDP, dark field image, and electron energy loss spectrum (EELS) (Fig. S12   ities with that of whiskers and hillocks from Sn and Sn alloy films [51,52], since all of these processes are strongly influenced by the need for releasing Li deposition-induced stresses.Let us consider the SEI shell of Li 2 CO 3 as an example.In the initial stage of Li deposition, a spherical Li particle was formed.Due to weak mechanical constraints from the ultrathin Li 2 CO 3 shell, the Li depositioninduced stresses were low in both the Li core and Li 2 CO 3 shell, such that they had a weak impact on the growth of the Li particle.With further Li deposition, continued expansion of the Li particle with concomitant thickening of the Li 2 CO 3 shell resulted in the buildup of large compressive stresses in the Li particle as well as large hoop tensile stresses in the Li 2 CO 3 shell.With the occasional occurrence of non-uniform growth, the high stresses in the Li core could result in creep flow [53] of deposited Li (Fig. 7a), given the fact that the room temperature of 295 K is about 0.65T m , where T m is the melting point (455 K) of Li.Such creep relaxation of Li depositioninduced stresses smoothed out the local non-uniform growth, resulting in a nearly spherical shape of the Li particle.On the other hand, the buildup of large hoop tensile stresses in the nanocrystalline/amorphous Li 2 CO 3 shell could cause the formation of defects and even cracks in the shell, so as to release the high stresses induced by continued Li deposition in both the Li core and Li 2 -CO 3 shell [51].
After crack formation in the Li 2 CO 3 shell, the rate of Li growth abruptly increased because the resistance to Li deposition from large compressive stresses in the Li core was released substantially, while the driving force for Li deposition from applied voltage was still maintained.The above in situ observations have revealed three distinct Li growth modes (Fig. 7b-d) resulting from the dynamic interplay of cracking and self-healing of the Li 2 CO 3 surface layer.As shown in Figs.2a-j and 7b, when the magnitude of applied voltage is high (<À1.0V), the growth of Li whiskers predominates, once a 3D ring crack forms in the Li 2 CO 3 surface layer.Because of the high voltage and resultant fast rate of Li deposition, the axial growth is preferably continued in the same cross-sectional region (marked by a red box in the right diagram of Fig. 7b) where the ring crack is initially located.The continued local growth is favored, because the side surface of this cross-sectional region is effectively kept unhealed and thus experiences the least mechanical constraint on axial growth.In addition, as the newly grown crosssectional layer of Li moves away from the ring crack region, the Li 2 -CO 3 film gradually grows on the surface of this layer and thus imposes an increasing mechanical constraint on its growth.As a result of continued axial growth in the same cross-sectional region, the growing Li whisker maintains a constant diameter and an unchanged shape of the top part (Figs.7b, 2j).This mechanism of axial growth is further confirmed by our in situ TEM observations, as to be discussed next.Hence, the high voltage and fast Li deposition rate play a critical role in the formation of Li whiskers.
When the applied voltage is in the intermediate range (À0.3 to À0.8 V) (Fig. 7c), the growth of Li deposits features several rounds of local extrusion and rounding of Li particles.In this case, the Li particle grows preferentially near the cracked region.Since the rate of Li deposition is lower than the case in Fig. 7b, the crack in the Li 2 CO 3 film has sufficient time to heal.Crack self-healing suppresses further local extrusion, resulting in local rounding of Li particles.Growth of a large particle usually involves several rounds of cracking and self-healing, accompanied with repeated processes of local extrusion and rounding.As schematically shown in Fig. 7c, when a local crack is formed in the Li 2 CO 3 film, Li growth is localized to the cracked region.Since the rate of Li deposition is relatively low, creep flow of Li in this region can smooth the local irregular shape.This smoothing process can be understood in terms of a capillary effect of reducing the local high curvatures of extruded regions [38].Moreover, the local crack in the Li 2 CO 3 film can be healed through reaction between Li and CO 2 .Hence, the intermediate voltage and associated intermediate Li deposition rate are responsible for the growth mode with intermittent local extrusion and rounding in the shape of Li particles.
When the applied voltage is in the low range of À0.1 to À0.2 V (Fig. 7d), the rate of Li deposition is low, so that the Li depositioninduced stress is low.As a result, no major cracks form in the Li 2 -CO 3 layer.Instead, the growing Li sphere is accommodated by the expanding Li 2 CO 3 shell.That is, the slow and uniform expansion of the growing Li sphere results in the low hoop tensile stress and accordingly the uniform expansion in the nanocrystalline/amor-   phous Li 2 CO 3 shell, the latter of which likely involves the elastic/plastic stretching as well as the concurrent growth of Li 2 CO 3 in the shell plane due to continuous reaction between Li and CO 2 .As a result, the expanding Li 2 CO 3 shell can maintain a nearly uniform thickness as the Li sphere grows.Hence, the low voltage and associated low Li deposition rate give rise to the uniform growth of spherical Li particles.
From a further careful tracking of cracked SEIs, we found that the continued growth of a Li whisker often occurred at a weak spot along its length where the SEI cracked.Such weak-spot growth contrasts with the generally recognized tip growth and bottom growth.As shown in Fig. 8, the time-resolved TEM observation of a growing Li whisker clearly reveals the weak-spot growth.Fig. 8a shows a representative low-magnification TEM image where local SEI cracking (indicated by a red arrow) just occurred at a middle cross section of a growing Li whisker.In the corre-sponding time series of high-magnification TEM images, it is seen that prior to SEI cracking (Fig. 8b), the Li whisker was covered by an SEI layer (in dark gray) with a uniform thickness of about 10.6 nm.Upon the onset of local SEI cracking (Fig. 8c), both the upper edge (marked by a blue arrow) and the lower edge (marked by a red arrow) of the cracked SEI film were clearly identified.During the subsequent growth of this Li whisker (Fig. 8d-k), the lower edge remained almost fixed, while the upper edge continuously moved upward along the growth direction.Obviously, the growth of the Li whisker occurred predominantly in between the two edges that became increasingly separated.Furthermore, the newly formed thin SEI layer (in dark gray) between the two edges exhibited a gradual decrease in thickness from the lower to the upper edge of the cracked SEI film (e.g., see Fig. 8g, k).Such thickness variation indicates the occurrence of continued axial growth of the whisker in its cross-sectional region close to the upper edge  of the cracked SEI film.As discussed earlier, the continued local growth at a fixed cross-sectional region is favored, because the side surface of this cross-sectional region is effectively kept unhealed and thus subjected to least mechanical constraints on the axial growth of the whisker.
For the whisker shown in Fig. 8a-k, we measured its axial length versus time, which exhibits a linear relationship, as shown by the blue line in Fig. 8l.Also shown in Fig. 8l is another linear curve measured from a different Li whisker, supporting the generality of the linear time law of growth.The growth rates of Li metal are 14.4 and 2.99 nm s À1 , respectively.This linear relationship indicates that the rate-limiting growth step of the Li whisker should involve some ''local" Li reaction instead of Li diffusion along its axial direction.Based on observations from Fig. 8a-k, the ''local" Li reaction can be reasonably attributed to the process of Li insertion in a cross section of the Li whisker as illustrated in Fig. 8m, n.In other words, during the axial growth of the Li whisker, Li insertion in a cross section should be effectively treated as a local reaction, while it likely involves the radial Li diffusion within this cross section.For a similar phenomenon of the axial growth of singlecrystal Sn whiskers [54], it has been suggested that Li insertion in a cross section occurred through the Bardeen-Herring mechanism of climb of a prismatic dislocation loop.Such a prismatic loop has the Burgers vector perpendicular to the cross section and its climb can result in the axial growth of the whisker.The loopclimbing process involves radial diffusion of atoms in the crosssectional plane as well as local insertion at the dislocation core.In the same vein, the axial growth of Li whiskers could occur via the Bardeen-Herring mechanism of climb of prismatic loops as well.The detailed atomic-scale process of Li insertion in the cross section of Li whiskers cannot be resolved from the present TEM experiment and warrants further studies via for example cryo-TEM imaging [49] in the future.Nevertheless, it is remarkable that by virtue of ETEM the growth of lithium dendrite was observed in real time at room temperature.We believe that the growth of a Li 2 -CO 3 shell on the surface of the lithium dendrite plays an essential role in protecting lithium from electron beam damage.Based on our experimental results, during discharge of the nanobattery in a CO 2 atmosphere, a Li 2 CO 3 shell formed on the surface of the Li dendrite, which is critical in stabilizing the Li whiskers under electron beam.Without Li 2 CO 3 protection, the deposited lithium was unstable and collapsed easily under electron irradiation [48].
The observation of cracks in the Li 2 CO 3 layer but not in the Li 2 O layer is attributed to the different physical, chemical and mechanical properties between the two materials.Based on our experimental results, the Li 2 O layer is thicker and looser than Li 2 CO 3 layer.Moreover, the mechanical properties of Li 2 O and Li 2 CO 3 are different.Billone et al. [55] reported that the Li 2 O is flexible and can buffers mechanical stress from lithium deposition.Our recent results show that Li 2 CO 3 is more brittle than Li 2 O, and the former exhibits brittle fracture characteristic.On the other hand, Li 2 O displays good plasticity with a ductility of almost 30%.We believe that it is the brittleness of Li 2 CO 3 and the ductility of Li 2 O that lead to the different mechanical response between the two materials during lithium deposition.
In a final note, we compare our in situ TEM results with that from practical cells.Lithium dendrite is hypothesized to be shot out from pinholes in an SEI film [14,16,17] due to stress buildup during lithium deposition.However, such hypothesis has never been demonstrated experimentally.Our experiment appears to be the closest evidence that one can achieve hitherto to demonstrate the possibility of such a dendrite growth mechanism.However, the difference in the dendrite growth between our in situ ETEM experiments and realistic cells is apparent, namely, in the latter, dendrite shoots out at the Sand's time when diffusion limited growth sets in Ref. [16]; while in the former, neither the Sand's time nor cationic exhaustion seems to be operating.The difference may have arised from the following factors: (1) realistic cells operate in a liquid electrolyte, while the in situ TEM cell operates in an all solid configuration, and the Li + ion diffusion kinetics and thermodynamics are entirely different between the two experiments; (2) The SEI in realistic cells comprises of both inorganic and organic components [18], while that in the in situ experiment consists of either Li 2 O and Li 2 CO 3 only.Nevertheless, because Li 2 O and Li 2 CO 3 are the major components of the SEI in realistic cells, we believe that our experiments capture the main mechanics, physics and chemistry of SEI, which has important implications in understanding SEI and related dendritic problems in lithium batteries.For example, we show that it is possible to tailor the lithium deposition morphologies by controlling voltages and the SEI constituents in our in situ cell, and similar strategies have been adopted in practical battery cell, namely by controlling the current density and the SEI composition, the dendrite growth can be mitigated.

Conclusion
Our in situ TEM experiments have revealed the dynamic interplay of SEI cracking, self-healing and Li deposition that can critically affect the morphology of Li deposits.Under high applied voltage, the high rate of Li deposition causes SEI cracking that dominates over SEI self-healing, leading to the axial growth of Li whiskers.At low applied voltage, the self-healing rate of cracks becomes sufficiently high compared to the Li deposition rate, leading to the growth of spherical Li particles.We have also demonstrated that it is possible to control the Li deposition morphology by varying SEI constituents and Li deposition rates.With an SEI film of Li 2 CO 3 , the growth of Li whisker or sphere can be controlled by applied voltage.For an SEI layer Li 2 O, the growth of Li sphere rather than whisker prevails, regardless of the magnitude of applied voltage.We have further found that the continued growth of a Li whisker can occur at a weak spot along its length due to local SEI cracking, which contrasts with the generally recognized tip growth or bottom growth of Li whiskers.Our results highlight the crucial role of applied voltage and SEI in controlling the morphology of Li deposits.Broadly, our work underscores the notion of tailoring Li deposition morphology and mitigating Li whisker growth by development of designer SEIs and electrolyte additives, as well as by controlling the operating voltage/current.
online) confirmed that the Li particles formed in O 2 gas consist of a Li core covered with a Li 2 O shell.The Li 2 O layer is usually thicker than the Li 2 CO 3 layer formed on the surface of Li particles, as shown in Fig. S13 (online).Hence, the dominant growth of spherical Li particles can be attributed to less cracking and faster self-healing in the SEI film of Li 2 O as opposed to Li 2 CO 3 .An in-depth understanding of these differences warrants further study of the electro-chemo-mechanical properties of Li 2 O and Li 2 CO 3 nanolayers in the future.Next we consider the effects of Li deposition-induced stress on Li deposit morphologies.By in situ observations of Li deposition under the mechanical constraint of Li 2 CO 3 or Li 2 O SEI films, we find that the growth of Li whiskers and particles shares striking similar-

Fig. 1 .
Fig. 1. (Color online) Schematic of the experimental set up.The experiment was carried out in an aberration corrected environmental transmission electron microscopy under a CO 2 atmosphere.

Fig. 2 .
Fig. 2. (Color online) In situ growth of an electroplated Li whisker at a high applied voltage in a CO 2 atmosphere.The applied voltages are À0.3 V in (a, b) and À1.0 V in (e, j), respectively.Upon applying a voltage at the CNT cathode, a spherical Li particle emerged at the three-phase point.(c-j) Time lapse TEM images showing the formation of a Li whisker.Arrowheads mark cracks.Arrows point out the whisker growth direction.(k-m) TEM bright field, electron diffraction pattern (EDP), and dark field images of a Li whisker.The dark-field image in (m) was obtained by choosing the sharp diffraction spot marked by a red circle in (l), demonstrating that the straight Li whisker is a single crystal.The EDP indicates the newly formed Li with a [011] zone axis growing along the h1 1 1i direction (l).

Fig. 3 .
Fig. 3. (Color online) In situ growth of an electroplated Li particle at a low applied voltage (À0.5 V) in a CO 2 atmosphere.(a-p) Time lapse TEM images showing four cycles of cracking and self-healing of the SEI shell of Li 2 CO 3 .The first to the fourth cycles are shown in (a-d), (e-h), (i-l) and (m-p), respectively.The arrowheads indicate the cracks formed, while the arrows indicate the direction of local expansion during the growth of this Li particle.

Fig. 4 .
Fig. 4. (Color online) In situ growth of Li sphere under a low applied voltage (À0.2 V) in a CO 2 atmosphere.(a-p) Time lapse TEM images showing the formation of an electroplated Li sphere covered with an SEI shell of Li 2 CO 3 .Cracks in the SEI shell were rarely observed during the growth of this sphere.Even if an incipient crack (marked by red and blue arrowheads in (f-g), (n)) formed, it healed quickly.

Fig. 5 .
Fig. 5. (Color online) In situ TEM studies of the transformation of Li whisker into Li sphere in a CO 2 atmosphere.The applied voltages are À3.0 V in (a-c), and À1.0 V in (d-h), respectively.(a-h) By decreasing the applied voltage, an initially formed Li whisker was transformed into a nearly spherical Li particle.Yellow dashed lines outline the shapes of Li whisker and particle.

Fig. 6 .
Fig. 6.In situ growth of Li sphere at a high applied voltage in an O 2 atmosphere.The applied voltages are À0.4 V in (a-e), and À1.5 V in (f-h), respectively.(a-h) Time lapse TEM images showing the formation of two spherical Li particles with an SEI shell of Li 2 O at a high applied voltage in an O 2 atmosphere.No cracks were observed during the whole process of Li deposition.

Fig. 7 .
Fig. 7. (Color online) Schematic illustration of three different modes of Li growth governed by the dynamic interplay of cracking/self-healing of SEI and Li deposition rate (controlled by applied voltage).(a) Schematic of the Li plating electrochemical device.(b) Growth of a Li whisker from an initial Li particle under a high applied voltage.A 3D ring crack forms in the SEI shell (with the upper and lower crack edges marked by red lines) of a Li particle (left).Cracking leads to directional growth of a Li whisker.(c) Growth of a Li particle with local extrusion under an intermediate applied voltage.(d) Growth of a spherical Li particle under a low applied voltage.An initially small particle (left) grows into a large particle.

Fig. 8 .
Fig. 8. (Color online) Time lapse TEM images showing the ''weak spot" growth mode of a Li whisker, which differs from the generally recognized tip or bottom growth mode.(a) A low-magnification TEM image showing the local cracking (marked by a red arrow) in the SEI shell of Li 2 CO 3 .(b-k) Time series of high-magnification TEM images showing the axial growth of the Li whisker between the upper edge (marked by the blue arrow) and lower edge (marked by the red arrow) of the cracked SEI film.(l) Measured length (L) versus time (t) for the Li whisker in (b-k) (blue circles) as well as for another Li whisker (pink triangles).The solid lines are linear fitting curves.(m) Schematic illustration showing the sequential processes of Li diffusion and insertion during the growth of a Li whisker, including Li supply from the Li metal anode, Li ion diffusion through the solid electrolyte Li 2 CO 3 layer covering the Li anode, Li ion diffusion through the SEI Li 2 CO 3 shell covering the Li whisker core, and Li insertion into a cross-sectional region of the Li whisker where the local SEI shell breaks.The last process is considered as the rate-limiting step of ''local" reaction during the axial growth of the Li whisker.(n) Schematic illustration of Li insertion in a cross section of the Li whisker through the Bardeen-Herring mechanism of climb of a prismatic dislocation loop.Such a prismatic loop has the Burgers vector b perpendicular to the cross section and its climb via atom insertion at the dislocation core can result in the axial growth of the whisker.