Synchronous Healing of Li Metal Anode via Asymmetrical Bidirectional Current.

The creation of Li metal anodes while minimizing dendrite growth is an important challenge for developing high-energy density batteries. Dendrites can originate from an inhomogeneous charge distribution or an irregular substrate, and often, the way to suppress dendrite growth is to avoid their formation altogether (ion-uniform mechanism over a shelf time). Herein, we propose a different route to eliminate dendrite formation, called an asymmetrical bidirectional current mode (ABCM) of charging, leading to a healable Li metal anode and resulting in a positive feedback cycle. This mode allows for a stable cyclic performance and suppresses dendrite formation effectively (while holding the polarization ∼27 mV for over 1,000 h), and provides a better result than suppressing Li dendrites via weakening of the Li dendrite (ion-uniform mechanism). These results indicate that ABCM may be a promising way to stabilize the Li anode of Li metal batteries, without any chemical/physical modification of the anode.

Introduction

Li metal is a promising candidate for the anode of next-generation high-energy-density battery (Deng et al., 2019, Liang et al., 2017, Lin et al., 2017, Zhang et al., 2020). A stable Li metal anode is a key point for the practical application of Li metal battery (LMB) (Cheng et al., 2017, Lu et al., 2014, Qiao et al., 2017, Zhang et al., 2018). Therefore, explorations on non-dendrite anode and reducing dead Li are eagerly required, such as high-modulus SEI film coating on Li metal (Liu et al., 2019a, Shi et al., 2018, Wang et al., 2017b), high-surface-energy substrate (Hou et al., 2019, Wang et al., 2017a), low local current density (Zhang et al., 2017), more reaction sites (Yang et al., 2015), cross-linking film (Zhao et al., 2019), and so on (Liang et al., 2019, Westover et al., 2019, Xu et al., 2019). However, once the uncontrollable dendrite occurs, the negative feedback cycle (SEI broken, dendrite worsen and so on) that follows would cause Li metal anode's failure quickly. To overcome such problem, healable anodes are rising recently to suppress inducing-dendrite issues, by the methods such as optimized charge distribution or thermodynamic nucleation (Huang et al., 2018, Hundekar et al., 2019, Zhu et al., 2019). Ding et al. proposed healable Li metal anode via Cs+ additive with electrostatic shield mechanism (Ding et al., 2013). Li et al. proposed healable anode via heating-induced nucleation affection (Li et al., 2018a). Meanwhile, an appropriate stress distribution also can lead to a healing behavior with a zipper-like-SEI mechanism (Wang et al., 2018a, Wang et al., 2018c).

As we know, the surface charge distribution of Li metal is highly affected by extra filed force, such as Lorentz force and external power supply (Li et al., 2018b, Shen et al., 2019, Wang et al., 2019). Therefore, we can regulate the charge distribution via programming charge-discharge mode (Liang et al., 2019, Yang et al., 2014). For example, inducing external alternating current could enhance the diffusion of Li ion and suppress dendrite growth (Chen et al., 2019b). Pulse current could suppress dendrite growth via a kinetic-based molecular mechanism (Li et al., 2017). In another work, sinusoidal ripple current with a shelve time could lead to a charge relaxation and relieve the dendrite (Zhang et al., 2019). All of these approaches are based on the concept of preventing dendrite before its occurrence; in contrast, the route for repairing dendrite (a healing route) is rarely reported. Herein, we designed two types of route to suppress Li dendrite growth. One is intermittent-deposition route by a shelve time for charge relaxation mode (STCRM), corresponding to a common suppressing route. Another is asymmetrical-deposition route by reverse current for dendrite dissolution, named as asymmetrical bidirectional current mode (ABCM), of which the shelve time is changed into an anti-direction current, to dissolve the fresh-growth Li dendrite. By this strategy, Li dendrite can be suppressed with a synchronous healing mechanism (Figures 1 and S1). Commonly, Li dendrite will be aggravated under the continuous constant current, whereas during STCRM process, an inhomogeneous distribution of Li ion is changed into homogenization along with a charge relaxation, resulting in a weakened dendrite. Although small-current-density way also can suppress the growth of Li dendrite, giving up the rate performance is incompatible (Zhang et al., 2016). In addition, small-current-density strategy will still meet the battery failure because of the continuously accumulated dendrite. A step further to perfection, using ABCM process, the new and mild dendrite can be corrected during anti-direction current. Then Li metal anode can be healed continually by a positive feedback mechanism.

Figure 1

The Schematic Diagram of Three Charging Modes of Li Metal Plating

(A) The profiles of the applied current of three charging modes at the same special capacity and current density.

(B) A schematic diagram of CM (chronoamperometry mode), STCRM, and ABCM effecting on the growth/suppression of Li dendrite.

This figure is related to Table S1 and Figure S1.

Result and Discussion

To prove this concept, first, an in situ electrochemical visual battery was taken to investigate the healing phenomena during mild dendrite state (detailed in Supplemental Information, Transparent Methods part), as shown in Figures 2, S2, and S3 and Videos S1, S2, S3, S4, S5, and S6 (Pang et al., 2017, Wood et al., 2016). Two types of dendrites (serious dendrites and weaken dendrites) can be found after their growth at 5 mA cm−2. Figure 2A, S3A, and S3F and Videos S1 and S5 show that the large dendrite cannot be dissolved during the discharge state. On the contrary, Figures 2B, S3B, and S3G and Videos S2 and S6 show fresh and mild Li dendrite can be dissolved by anti-direction current. In addition, long-time Li plating behaviors at conditions of CM (chronoamperometry mode) and ABCM have been vividly exhibited (Figures 2C–2F and S3C–S3E and Videos S3 and S4). Under the condition of CM, the surface morphology is running to disorder and rough with a lot of dendrites (Figure 2C and Video S3). Oppositely, the surface morphology is kept smooth with micron-level Li sphere stacking closely by ABCM (Figure 2D and Video S4). Even at a special capacity of 3 mAh cm−2, the surface of the ABCM sample is still smooth (Figure 2E), whereas dendrites are very serious and chaotically distributed for CM sample (Figures 2F and S3C–S3E). Commonly, vertical columnar dendrite would be easily corrected as mild dendrite, whereas branch dendrite would be difficult to be repaired as serious dendrite (Figure S3H and detailed in Supplemental Information). Noting that we use visual batteries only to prove our healing mechanism, the real batteries such as coin batteries and pouch cells are different from this type of battery (pressure, the mount of electrolyte, separator, and so on), and the data and discussion based on coin cell are shown in the following part. Moreover, to obtain a more intuitive experimental phenomenon to prove our concept, we have discussed different current density for lithium deposition and dissolution. Meanwhile, small dendrites can easily be corrected by tuning current density, which is proved by both modeling work and experiment (Yang et al., 2019). Additionally, it has been reported that 100% Li DoD (Depth of Discharge) in Li/S battery can lead to a better surface morphology than 36%, indicating that a reasonable surface can benefit the cyclic performance (100% discharge state means a smooth current collector, such as copper foil) (Mikhaylik et al., 2010). Therefore, healable Li metal anode can be obtained logically by corrected Li dendrite timely (holding a smooth surface always) using anti-direction current.

Figure 2

In Situ Electrochemical Visual Battery for Tracking the Dendrite Dissolution

(A) Serious dendrites-related irreparable surface state using Li/Li system.

(B) Mild and fresh dendrites-related repairable surface and dendrite-correcting course using Li/Li system.

(C) Plating Li by CM using Li/Cu system (5 mA cm−2).

(D) Plating Li by ABCM using Li/Cu system (plating Li for 31 s and 5 mA cm−2, dissolving dendrite for 5 s and 1 mA cm−2).

(E) Plating Li by ABCM at 3 mAh cm−2 using Li/Cu system (plating Li for 31 s and 5 mA cm−2, dissolving dendrite for 5 s and 1 mA cm−2).

(F) Plating Li by CM at 3 mAh cm−2 using Li/Cu system (5 mA cm−2).

The scale bar is 100 μm. This figure is related to Videos S1, S2, S3, and S4 and Figures S2 and S3.

Figures S4A–S4F show polarizations of three type charge modes at pre-deposition Li/Li symmetrical cell system for 100 cycles (Li/Cu cell initially). All of them exhibit ∼19 mV polarization at initial state. The electrochemical performance of CM sample became worse suddenly at 180 h (Figures S4A and B), and the polarization reached around 25 mV at the 100th cycle. On the contrary, Figures S4C–S4F indicate that STCRM and ABCM can keep a low and stable polarization around 15 mV. Magnified voltage profile at the 100th cycle demonstrates that CM leads a concave curve and the highest polarization (Figure S4B). Meanwhile, STCRM shows a pulse-shape voltage profile and a stable polarization, where the voltage in shelve-time of every charge/discharge step is above/below zero in order, correspondingly. In contrast, during charge/discharge progress of ABCM, every inverse voltage for correcting dendrite is below/above zero, respectively (Figure S4F). As intermittent current and bidirectional current modes cost more time under the same current density/cyclic capacity (Figures S4B, S4D, and S4F), herein, we also investigate these modes under the condition of same time and same cycle capacity, noting that ABCM and STCRM own different rates at the same current density and the same cyclic capacity (caused by anti-current time and shelved time, which cost more time, as shown in Figure 1A) from CM. To give a fair comparison, we take the same rate to evaluate these three modes in Figures S5–S7. All of the three modes have the same cyclic capacity at the same charge/discharge time. Under such conditions, STCRM and ABCM need a higher current density than CM at the normal charging and discharging process, as shown in Figure S1B. And the results show that, in spite of enlarging current density, STCRM and ABCM also exhibit better performance and more smooth deposited morphology.

In addition, in the view of initial nucleation affection, Li/carbon coating copper batteries were also investigated (detailed in Supplemental Information, Transparent Methods part) (Figures 3A and S8) (Chen et al., 2019a, Liu et al., 2019b, Lu et al., 2017, Pei et al., 2017). At the same pre-deposited condition, both STCRM and ABCM can keep long stable polarization for more than 1,000 h Figures 3A and S8 show both can tailor initial polarization mildly, and after only 40 cycles, the polarization reduced below 30 mV. As experiments went on, the polarization of STCRM only lasted about 600 h below 30 mV and then presented a significant increase. It is noteworthy that ABCM can hold the polarization around 27 mV for over 1,000 h. To give a more obvious comparison, Figure S8 gives voltage profiles of 200th, 600th, and 1,000th h for STCRM and ABCM; it shows ABCM owns an obvious advantage for long cycles (37 mV [STCRM] versus 28 mV [ABCM] of the cycle at 1,000th h). It suggests that optimized Li ion distribution and timely corrected surface state can successfully obtain a stable reduced polarization, which reflected in suppressing dendrite (Liu et al., 2018, Salvatierra et al., 2018). Especially, the ABCM route can lead to a more stable performance compared with the weakened dendrite mechanism, because the weakened dendrite mechanism will reach the critical failure point while aggregating lots of little dendrite within the long-term cycle. Furthermore, Figure S9 shows the sample at the ABCM condition of A3 in Table S1 also owns optimized performance, lasting for more than 800 h.

Figure 3

The Polarization, Surface Morphology and Mechanism of Li Plating/Stripping by These Three Charging Modes

(A) The 1,000-h polarization of ABCM (B2 in Table S1).

(B) The 200-h polarization of CM-h/STCRM-h/ABCM-h (as black, green, and orange curves; C6, C8, and C9 in Table S1, respectively) at the same cyclic time and capacity (3 mAh cm−2) by using ether-based electrolyte.

(C) A schematic diagram of Coulombic Efficiency for CM and ABCM and a mechanism of semi-dead Li.

(D) The 150-h polarization of CM/STCRM/ABCM at the same current density (0.5 mA cm−2) and capacity (0.5 mAh cm−2) by using carbonate-based electrolyte.

(E) SEM of CM after 150-h symmetric cycling.

(F) SEM of STCRM after 150-h symmetric cycling.

(G) SEM of ABCM after 150-h symmetric cycling.

See E1–E3 in Table S1, correspondingly. This figure is related to Figures S8 and S10 and Table S1.

Additionally, we also investigated the polarization of these three modes for Li/Li cells (commercial lithium foil) at the same and 3-fold currents (see C1–C6 and C8–C9 in Table S1). Figure 3B shows the comparison at the same rate, and there is an obvious healing caused depolarization phenomenon during the early stage of STCRM and ABCM, and the stable polarization below 50 mV could last more than 200 h. Although CM-h (refer to C6 in Table S1) owns the lowest initial polarization resulting from the lowest initial current density (CM [10 mV], STCRM [20 mV], and ABCM [18 mV], as shown in Figure S10), CM-h reveals an increasing polarization suddenly and continuously, which is affected by the nucleation and surface electron distribution. After 90-h loops, CM-h presents the highest polarization in these three samples. Meanwhile, the STCRM-h (refer to C8 in Table S1) exhibits a stable and tailoring polarization and the polarization is around 60 mV. In contrast, ABCM-h (refer to C9 in Table S1) shows the best cyclic performance, where a tailoring polarization becomes smaller and smaller gradually and stable at 35 mV. Moreover, Figure S11 illustrates ABCM also owns obvious advantage than CM and STCRM at 1 mAh cm−2. Based on those electrochemical results, it is clear that the STCRM and ABCM surpass the traditional CM in the aspect of stability, duration, and smaller polarization, whereas ABCM offers the lowest polarization and the best stability. Generally, all of symmetrical cell systems (Li foil/Li foil or deposited Li/Li foil) show ABCM can always obtain the most stable polarization.

In addition, in carbonate-based electrolyte, CM exhibits a continuous increasing polarization and a sudden reducing at 120 h (short circuit); STCRM likewise shows a continuous increasing polarization, but the polarization is still smaller than CM. Meanwhile, ABCM owns the lowest and most stable polarization of ∼60 mV (Figure 3D). For the pre-deposited Li metal (13 mAh cm−2), CM cannot last more than 75 h at 1 mA cm−2 and 1 mAh cm−2, in contrast with the stable cycle performance of ABCM (Figure S12). In general, ABCM exhibits the best performances not only in ether-based electrolyte but also in carbonate-based electrolyte.

Figure S13 gives voltage-time curves of Li/Cu cell for CM and ABCM; each cell was deposited with the same amount of Li on carbon-coated copper, then after 10 loops of CM or ABCM route (the same as A1 or A3, respectively), the Li on carbon-coated copper is striped by electrochemical dissolution under 2V, then C.E. (Coulombic Efficiency) can be calculated and labeled on the corresponding cycle (Adams et al., 2018, Fan et al., 2018). The C.E. of CM sample is lower than that of ABCM, the C.E.s of five loops for CM are 96.3%, 97.6%, 77.2%, 95.1%, and 90.0% in order. Correspondingly, C.E.s of five loops for ABCM are 95.2%, 97.2%, 98.0%, 97.6%, and 94.4%, respectively (Figure S13). Hypothetically, pre-deposited Li metal owns 100% C.E., thereof, for 50 cycles of these three modes, the average C.E.s of CM/STCRM/ABCM can be gained as 91.24/95.60/96.28%, respectively (Figures S13 and S14). It means both STCRM and ABCM exhibit optimized reversible reaction and ABCM owns the best reversibility. Meanwhile, Figure S13B demonstrates that, during the final striping course, the voltage curve of CM is disorder and churning. On the contrary, ABCM leads to a more stable process (Figure S13D). Possibly, CM has more side reaction or more dead Li/semi-dead Li. The dead Li is electrically isolated from the substrate and semi-dead Li means that Li metal contacts with Li metal by some shared SEI film (Figure 3C); meanwhile a recent report has proved semi-dead Li real existed (Fang et al., 2019). The bulk Li source on substrate can be stripped smoothly; however, the semi-dead Li induces more electronic resistance due to the shared SEI film. Along with increasing stripping voltage, the resistance of shared SEI film may be breakdown, resulting in an inauthentic C.E. data (Figure S13B). Besides, carbonate-based electrolyte has also been investigated for C.E., indicating that ABCM exhibits better performance than CM (Figure S15). Although ABCM can optimize C.E. performance in these two types of electrolyte, the intrinsic SEI is still a dominant issue for the protection and effective utilization of lithium metal anode.

Correspondingly, Figures 3E–3G, 4A–4G, and S16A–S16E give the morphology, chemical component of SEI film, and interfacial impedance. In the carbonate-based electrolyte, scanning electron microscopy (SEM) shows that ABCM exhibits a dendrite-free surface, in contrast with accumulation of Li pieces (STCRM) and serious dendrites (CM) (Figures 3E–3G). Besides we also investigate the surface morphology in ether-based electrolyte. Obviously, the morphology of the CM sample is rough and chaotic (Figures 4A and S16A). Figures 4B, 4C, S16B, and S16C illustrate a flat surface and the Li anode is assembled via a uniform zoned distribution, and the zoned areas of STCRM-100 and ABCM-100 are around 40 and 200 μm2 (the average particle size distribution is 6.4 μm [STCRM-100] and 13 μm [ABCM-100], as shown in Figures S18A and S18B), respectively. Figures 4D and S16D present the surface morphology of STCRM after 1,000-h cycles. Although the polarization in STCRM is a little worse (Figure S8A), the surface morphology exhibits obvious roughness and pulverization (Figures 4D and S16D). At the same time, ABCM (Figures 4E and S16E) still exhibits a free-dendrite surface morphology, indicating the healable Li metal anode using corrected dendrite mechanism is more advanced and practical than a weakened dendrite mechanism (STCRM) during long-term cycles, and this result highly proves our concept (Figure 1B). Furthermore, X-ray photoelectron spectroscopy (XPS) was taken to analyze the differences between the three modes with carbonate-based electrolyte. We focus on F element, for it is stable to air and can reveal the information of the top-layer SEI. Herein, F1s fitting data are shown in Figures S17A–S17C: CM owns the obvious highest concentration of LiF peak (at ∼685.1 eV, 44.8%) in contrast with ABCM (25.1%) and STCRM (43.2%) (Liu et al., 2019c). As we know, the surface of SEI film is mainly constructed by organic fragment. The more the Li dendrite, the more the consumption of electrolyte and the more the formation of SEI, therefore, leading to a high concentration of LiF. Based on this, the XPS data can tell that the ABCM produces the fewest dendrites. After Ar ion etching (∼16 nm deep), information of the bottom SEI film has been investigated. The organic layer has been eliminated completely, because the peaks of organic component in F1s C1s O1s are not observed (Figures S17D–S17H) (Eshkenazi et al., 2004, Wang et al., 2018b). Therefore, Li 1s peak fitting data in Figure 4F give the real information of the inorganic layer, which is near the Li metal. The peak at ∼54.5 eV stems from Li2O2 and LiOH, which is an unstable state (Lu et al., 2008, Nasybulin et al., 2013). By contrast, the peaks at around 53.7, 55.1, and 55.9 eV are originated from Li2O, Li2CO3, and LiF, respectively (Edström et al., 2006, Liu et al., 2019a). Obviously, the peak at ∼54.5 eV in ABCM is the weakest, indicating the fewest broken SEI in ABCM route, which can be attributed to that very few new Li metal (dendrite) contacted with electrolyte by this route. The surface of ABCM-1000 (refer to B2 in Table S1) is still maintained at an average particle size of ∼10 μm (Figure S18C), in contrast with a disorder surface morphology of STCRM-1000 (refer to B1 in Table S1). Generally, there is rare dendrite by ABCM route, which is not serious for the battery system; by contrast, lots of dendrites and dead Li/semi-dead Li make battery deteriorative by the CM route (Chen et al., 2017).

Figure 4

Surface Morphology and Impedance of Cyclic Li Metal Battery via These Three Type Modes

(A–E) are SEM of CM-100, STCRM-100, ABCM-100, STCRM-1000, and ABCM-1000, respectively.

(F) Peak fitting of Li 1 s after Ar ion etch for ABCM, STCRM, and CM.

(G) The calculated SEI-film impedances of these EIS data. The illustration in Figure 4G shows the impedances of three samples, CM-200, STCRM-200, and ABCM-200, mean cycling for 200 cycles, and STCRM-250 and ABCM-250 mean going on cycling to 250 cycles (3 mA/cm−2). See A10–A12 in Table S1.

(H) Rate capacity and long-term stability of CM, STCRM, and ABCM at 0.4 C, 1 C, 2 C, 6 C, 10 C (at the rate stage) and 10 C (at the long cycle stage) for Li/LiCoO2 batteries. (The full circle stands for the specific discharge capacity, and the half-filled squares are the Coulombic Efficiency).

The scale bars in (A–E) are 20 μm. This figure is related to Figures S4, S16, S18, S19, and S22 and Table S1.

Figure 4G gives electrochemical impedance spectra (EIS) of these routes using higher rate performance (A10–12 as shown in Table S1); the arc curve in the illustration means the impedance of SEI film (Bieker et al., 2015). Therefore, CM-200 (refer to A10 in Table S1) exhibits the highest interface impedance. Simultaneously, ABCM-200/250 (refer to A12 in Table S1) exhibits excellent impedance at about 0.69 and 0.96 ohm, respectively. In addition, side reaction and dendrite would increase the interface impedance from electrolyte and electrode inherent character, named Rs (Figure S19) (Wang et al., 2018a). Obviously, CM-200 and STCRM-200/250 (refer to A11 in Table S1) exhibit higher Rs than ABCM 200/250. Figures S20 and S21 also give EIS data corresponding to Figures 4A–4E, where ABCM exhibits absolute advantage in long-term cycle and some advantage in short cycle, in accordance with the conclusion that traditional route (STCRM) can stabilize the Li metal anode via weakening dendrite; however, battery of STCRM will fail when dendrites accumulate too much. All of these results illustrate the ABCM route can provide a stable and long-life Li metal anode. These EIS data indicate ABCM can effectively heal dendrites without more electrolyte consumption (low Rs) and thicker SEI film (Low Rsei). Obviously, ABCM in ether electrolyte can get better performance than in carbonate-based electrolyte. Thereof, this technology is an optimized route and can reach stable Li metal anode with adaptive SEI film, showing a promise to reach the real practical application. Furthermore, full cells (Li/LCO battery) are taken to investigate the rate and duration of these three modes (Figures 4H, S22A, and S22B). By contrast, ABCM demonstrates the best performance at all current densities from 0.4 to 10 C and CM exhibits the worst performance. After 50 loops, the ABCM sample continues to cycle up to 500 cycles at 10 C and still exhibits ∼90 mAh g−1 capacity. The 500 cycles performances of CM, STCRM, and ABCM at 10 C are also shown in Figure S22. Although the average C.E.s are similar (∼99.7%), ABCM has the smallest fluctuation and exhibits the highest capacity retention as 83.9%, in contrast to 79.9% and 66.2% for STCRM and CM, respectively.

Moreover, Figure S23 shows that different Li plating time can affect the polarization, and Figure S24 shows that different dendrite correcting time can also directly affect the polarization. Therefore, we also should consider that irreparable surface can highly dominate the cyclic performance (Figure 1B). Meanwhile, different plating/dissolving times can induce different states (Figure S25): for ABCM-a, the dendrite cannot be repaired under the condition of C7 in Table S1, which results in sustained increasing polarization; however, for ABCM-b, under the condition of C9 in Table S1, the dendrite can be repaired and results in a healable polarization. Therefore, correcting dendrites timely is necessary in the ABCM process, which further proves our view of correcting dendrite.

Conclusion

Herein a healing route for suppressing Li dendrite was investigated, which corrects mild dendrite timely via reverse current. This route can effectively suppress Li dendrite, in contrast with traditional suppressing route, e.g., STCRM. ABCM exhibits a more promising way to obtain stable and long-life Li anode, because of eliminating Li dendrite timely and avoiding the positive feedback cycle, avoiding the agminated small dendrites reaching a critical failure point of battery by STCRM. In general, different from conventional suppressing Li dendrite, e.g., SEI film, nucleation route, and so on, our route puts forward another way (corrected dendrite timely for healing) to suppress dendrite. Our work indicates the improved charge protocol is also important for stable Li metal battery (different charge protocol is widely used in Li ion battery for increasing rate performance) (Alcaraz et al., 2017). And, the idea of correcting dendrite before the irreparable deterioration is proved as a promising way to harvest stable and dendrite-free Li metal anode. This viewpoint can be extended to other alkali metal electrodes, such as Na, K, and Zn.

Limitations of the Study

Our charging technique contains an anti-direction current for a short time regularly during charge, which causes a small energy loss of the battery. This loss of energy is quite little but should not be neglected (less than 15% in our work). Such energy loss is a secondary issue in contrast to the safety of the Li metal anode; however, efforts are still needed to minimize this energy loss in the further work.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.