Screening of salts and solvents
At the onset of SEI formation, the solvent sheet with Li+ will be reduced by the electrons accumulated in the vicinity of Li metal surface, resulting in the SEI components. To control the SEI composition requires that the preferred interface species should have a higher reduction potential as compared to the main solvents, which is equivalent to a higher electron affinity or a lower lowest unoccupied molecular orbital (LUMO)30. According to the LUMO energy obtained from density functional theory (DFT) simulations, shown in Fig. 1a and Supplementary Table 1, FDMA exhibits the lowest LUMO among all solvents studied, demonstrating a high electron affinity, which can be expected to decompose first and hence dominate the SEI formation. LiNO3, as a N-donating additive, has been widely used in ether-based electrolytes31,32,33, however, it is practically insoluble in carbonate solvents (<10−5 g mL−1) making this a challenging strategy for practical application21,34,35. Although FDMA has a comparable HOMO-LUMO gap to LiNO3, FDMA has the advantage of being highly soluble in both ether and carbonate solvents, providing a more practical route to introduce N-containing components in the SEI. Since FDMA has a relatively high HOMO level, the utilization of FEC as co-solvent with the lowest HOMO level among the widely used solvents is necessary to enhance the high-voltage stability towards the cathode by means of the formation of LiF17,19. Considering the Li salts, LiPF6 has the advantage of having a large HOMO-LUMO gap, however, FEC is found to be thermally instable in LiPF6-based electrolytes, which will trigger the generation of unwanted HF and various acids. These acids will cause significant dissolution of transition metal ions into the electrolytes, and lead to the serious degradation of cathode materials36,37. Other frequently used salts, having a similar electrochemical window, are LiTFSI and LiFSI, of which LiTFSI is reported to be more stable towards Li metal due to the stable –CF3 group38. According to the discussion above, we prepared 1 M LiTFSI in a solvent mixture of FDMA/FEC (1:1 by volume) as an electrolyte and explored its properties in Li-metal batteries.
To study the role of the FDMA in the SEI formation, possible decomposition mechanisms of FDMA were considered based on reaction energy calculations. Three reactions, including no Li+/electron attack, single Li+/electron reactions, and two Li+/electron reactions, are considered in combination with the five possible bond-breaking reactions as shown in Supplementary Fig. 1. The thermodynamically most likely first step of the FDMA decomposition is that two Li+ and two electrons attack the N–CH3 bond to form CH3Li and a N-containing species. In this amide molecule, the lone pair electron of the amino nitrogen will be conjugated with the π electron from the carbonyl group, in which the electron cloud on the nitrogen is less dense and therefore prone to accept an electron. As a result, we propose a possible three-step decomposition mechanism as shown in Fig. 1b and Supplementary Fig. 2. It is worth noticing that the decomposition of FDMA will produce Li3N in the second step, and that the final decomposition products are small molecule compounds which are unlikely to undergo further reaction. Furthermore, we calculated the HOMO-LUMO window of DMA and DMF for comparison as shown in Supplementary Fig. 3. The result shows that the introduction of –CF3 group reduces both the LUMO and HOMO levels, enabling FDMA to participate in the formation of SEI layers at higher potentials. Meanwhile, the stronger electron-withdrawing –CF3 group, as compared to that of –H and –CH3 groups, generates a repulsive force towards the C=O dipole, which contributes to an improved oxidation stability39.
Li||Cu cells with different electrolytes were assembled to investigate the Li plating/stripping upon electrochemical cycling as shown in Fig. 2a–c. Clearly, the as-prepared amide electrolyte, 1 M LiTFSI-FDMA/FEC (1:1 in volume), exhibits a significantly enhanced Coulombic efficiency and cycling stability as compared to that of the conventional 1 M LiPF6-EC/DMC (1:1 in volume) electrolyte and the reported performance of 1 M LiPF6-FEC/DMC (1:1 in volume)19. The average Coulombic efficiency of the Li plating/stripping reaches ~99.3% for the as-prepared electrolyte up to 100 cycles, which is a major improvement compared to conventional EC-based electrolytes as shown in Fig. 2c. The high reversibility of the Li plating/stripping sets in after several initial cycles, which is competitive to that observed in a concentrated fluorinated electrolyte (97.7% for the first 100 cycles)17. The synergy of the components in improving the cycling stability of the as-prepared amide electrolyte is confirmed by comparing the Coulombic efficiency of Li||Cu cells with different solvents and Li salts, as shown in Supplementary Fig. 4. In addition, the present amide electrolyte generates smaller overpotentials of ~65 mV against the highly fluorinated electrolyte (e.g., ~110 mV in 7 M LiFSI-FEC)17,19. The difference in cycling stability and overpotential is also observed in the symmetric Li-metal cells, especially at higher current densities, see Supplementary Figs. 5–8, where the amide electrolyte exhibits an overpotential of only ~10 mV and ~20 mV at a current density of 1 and 3 mA cm−2 for more than 1000 and 900 h, respectively, outperforming the EC-based electrolyte. Furthermore, the cycling stability at a high current density of 5.0 mA cm−2, and the rate test from 0.5 to 8 mA cm−2 further demonstrate the competitive reaction kinetics of this amide electrolyte (Supplementary Figs. 7 and 8). The high reversibility and remarkable cycling stability of the Li plating/stripping can be attributed to the interface chemistry that leads to a SEI layer in this electrolyte having the desired properties. A noticeable difference is found for the electrolytes after 100 cycles as shown in Supplementary Fig. 9, where the surface of Li metal anode in the EC-based electrolyte turns brown while in the present electrolyte retains its original color. Furthermore, electrochemical impedance spectroscopy is carried out on the Li||Li symmetric cells as shown in Supplementary Fig. 10, where the amide-based electrolyte exhibits a lower electrochemical impedance compared to the EC-based electrolyte, consistent with the above analysis.
The oxidation stability of various electrolytes was evaluated using linear sweep voltammetry (LSV) on Li||stainless-steel cells. The EC-based electrolyte shows a lower oxidation potential, as evidenced by a rapid increase in current above ~4.1 V as shown in Fig. 2d, while for the present amide-based electrolyte, the oxidation sets in at much higher potential. Cyclic voltammetry (CV) test of Li||Cu cells were performed to study the reductive stability as shown in Fig. 2e, f and Supplementary Fig. 11. It is observed that FDMA has a slightly higher reduction potential than that of EC and FEC. After several initial cycles, the reduction peaks disappear as shown in Supplementary Fig. 12, indicating that FDMA and FEC are only decomposed during the initial cycles.
Evolution of morphology for Li-metal anode
To investigate the Li deposition morphology in different electrolytes, ex situ scanning electron microscopy (SEM) and in situ optical microscopy are employed, the results of which as shown in Fig. 3a–d. After 100 cycles, the porous structure with needle-like dendritic Li for a few hundred nanometers is observed on the surface of Li metal anode after cycling in the EC-based electrolyte, as shown in Fig. 3a. In contrast, the as-prepared amide-based electrolyte leads to a compact aggregate of granular Li metal particles with sizes in the range of a few microns, as shown in Fig. 3b. The denser Li deposition will lead to less exposure of fresh electrolyte to the plated Li metal, reducing the detrimental decomposition reactions, resulting in a higher cycling stability. Furthermore, a dedicated cell is designed to monitor the Li plating/stripping in real-time, in situ, with an optical microscope. The images in Fig. 3c, d are taken at specific stages of cycling for the two different electrolytes and the corresponding movies are provided as Supplementary Movies 1 and 2, respectively. At first, the thickness evolution of the pristine Li metal was monitored during stripping at 1 mAh cm−2, where a slight decrease is observed for both electrolytes. Upon subsequent Li plating, protrusions appear on the surface of the Li metal electrodes in both electrolyte systems. However, in the EC-based electrolyte, uneven Li deposition is observed resulting in a porous morphology as observed from the side view. In contrast, in the amide electrolyte, a more homogeneous nucleation leads to denser Li deposition as shown in Fig. 3d. Upon subsequent stripping, almost all the Li deposits disappear in amide electrolyte, whereas a large amount of residual Li remains on the surface in the presence of the EC-based electrolyte. The corresponding evolution of Li-metal thickness during plating and stripping in both electrolytes, shown in Fig. 3e, confirms the denser and more uniform Li deposition in the amide electrolyte.
Interface between anode and electrolyte
To determine the interface decomposition species, X-ray photoelectron spectroscopy (XPS) analysis is performed on the cycled Li metal anodes for both the amide-based electrolyte (Fig. 4a) and the conventional EC-based electrolyte (Fig. 4b), where the survey spectra is shown in Supplementary Fig. 13. In the C 1s spectra, both electrolytes present the commonly observed species, including C–C/C–H (~284.8 eV), C–O (~285.7 eV), CO32− (~289.3 eV) as well as poly(CO3) (~290.3 eV)40,41. In contrast, for the amide electrolyte, two additional peaks appear that can be attributed to C–SOx (~287.6 eV)18 and CFx (~292.9 eV)42,43, most likely resulting from LiTFSI or FDMA. The existence of CFx (~687.6 eV) is further supported by the F 1s spectra, and the other peak at ~685 eV corresponds to LiF, which suggests the preferred reaction of FEC with Li metal44,45. For the EC-based electrolyte, the two major peaks at ~687 and 685 eV can be attributed to fluorophosphates (POxFy) and LiF, most likely generated via the reactions of PF5 and/or PF6 groups with Li metal, which testify the participation of LiPF6 in the SEI formation. For the Li 1s spectrum of the EC-based electrolyte, the peak in the range of ~53–56 eV can be deconvoluted into three components centered at ~55.7 eV, ~54.3 eV and ~53.2 eV, which can be assigned to LiF, Li2CO3 and Li2O, respectively46. For the amide system, two more peaks appear in the Li 1s spectrum, centered at ~58.1 eV and ~55.7 eV, which can be assigned to two N-containing species, Li–N–C and LixN, originating from the decomposition of FDMA and LiTFSI. From the results, the relative amount of Li-containing species at different depths is determined as shown in Fig. 4c. For both systems LiF accounts for a large portion of the SEI, where for the amide system, Li–N species (including Li–N–C and LixN) dominate the SEI chemistry. These Li–N species can be identified more clearly in the N 1s spectrum, where the three deconvoluted contributions (~402.1, ~400.4 and ~398.7 eV) can be assigned to Li–N–C, N–SOx and LixN, respectively42,47,48. After sputtering, the LixN signal becomes more apparent, along with a decrease of the Li–N–C signal, which can be explained by the different reduction products of FDMA, depending on the amount of reacting Li+ and e− as studied by the DFT calculations. In the inner part of the SEI, where electrons are able to reduce FDMA, it is more favorable to form LixN, while the poor electronic conductivity of the SEI makes that only the first decomposition products form at the outer part of the SEI.
Evolution of Li-metal density during plating and stripping
Operando neutron depth profiling (NDP) allows non-invasive monitoring of the spatial and temporal distribution of Li during electrochemical plating/stripping, providing direct visualization of the Li+ transport and irreversible processes49,50. The schematic setup of the operando NDP measurement and the corresponding principle are shown in Fig. 5a. Through the neutron capture reaction of the 6Li+ isotope (natural abundance 7.5%), two charged particles with a well-defined energy are produced, 4He2+ (Ek = 2044 keV) and 3H+ (Ek = 2727 keV). Due to the lower charge of the 3H+ particles, these are able to exit the pouch cells, whereas the stopping power towards 4He2+ is too large to penetrate the current collector49. Based on calculating the energy loss of the 3H+ particles, measured in the detector positioned perpendicular to the electrode, the 6Li+ density depth profile can be monitored during battery operation as shown in Fig. 5b. With a sub-micron depth resolution, and time resolution in the order of minutes, Operando NDP provides the unique opportunity to monitor Li plating/stripping quantitatively, under realistic electrochemical conditions.
Cycled at a current density of 1 mA cm–2 with a capacity of 1 mAh cm–2, the evolution of Li+ plating/stripping density vs. time are obtained from operando NDP during the first cycle in Li||Cu pouch cells for the 1 M LiTFSI-FEC/FDMA (Fig. 5c, d) and 1 M LiPF6-EC/DMC (Fig. 5f, g) electrolytes, respectively. As demonstrated by Fig. 5d, a denser and thinner Li deposition is observed for the amide-based electrolyte. In comparison, a much less compact and thicker Li deposit is observed for the EC-based electrolyte, extending to 18 μm into the electrolyte. These results are in good agreement with the ex situ SEM and in situ optical microscopy measurements in Fig. 3. It’s worth mentioning that the NDP measurements provide a more precise picture of the Li deposition as compared to the in situ optical microscopy, because for the latter there is a large space between the Li-metal and separator to get a clear observation of the Li-deposits, which is different to the practical cell. In the EC-based electrolyte, the Li density profile in Fig. 5g demonstrates a clear asymmetry comparing plating and stripping, consistent with the optical analysis. This phenomenon is more clearly visualized in the time derivative of the Li density, shown in Fig. 5h, representing the depth-resolved plating/stripping activity. The Li-metal stripping in the EC electrolyte is homogeneously distributed over the thickness of the deposit, whereas the stripping activity in the amide electrolyte moves back symmetrically, as compared to the plating, to the current collector. Homogeneous stripping will result in a porous Li-metal morphology, which promotes the formation of a high surface area and of Li-metal domains that are disconnected from the Cu current collector, thus resulting in the formation of “dead” Li. In contrast, stripping from the top of the deposit, as observed for the amide electrolyte in Fig. 5e and Supplementary Fig. 14, suggests a much more reversible stripping mechanism. Also, at higher and lower current densities, 0.5 and 5 mA cm−2, the evolution of the Li density still keeps reversible during plating and stripping using the amide electrolyte as shown in Supplementary Fig. 15 and Supplementary Fig. 16, respectively. The ability of NDP to quantitatively monitor the amount of Li metal on the current collector allows us to determine the Li efficiency, defined as the ratio between the stripped and the plated Li mass. This provides complementary information to the electron efficiency (Coulombic efficiency) where the difference quantifies the amount of irreversible reactions that do not involve Li-ion transfer, such as direct electrolyte reduction and chemical dissolution of Li from the SEI. Comparison between the Li-mass evolution in different electrolytes is shown in Supplementary Fig. 17, the amide electrolyte shows a better Li-efficiency and a higher Li mass after plating, indicating a lower amount of parasitic reduction reactions.
The origin of the top-down stripping and/or more reversible plating/stripping can be rationalized by the schematic representation in Fig. 5i, j. According to the XPS results, the SEI formed in the as-prepared electrolyte contains more ionic conductive components, which can be responsible for the denser Li metal deposits observed with operando NDP, as well as with SEM (Fig. 3b) and in situ optical microscopy measurements (Fig. 3d). Compared to the more porous Li metal/SEI morphology in the conventional electrolyte, compact Li metal deposition will allow more facile electron transport from the surface of the deposits to the Cu current collector, which will facilitate stripping from the top. In contrast, a more porous deposition will compromise electron transport and facilitate ion transport, resulting in relatively homogenous stripping. The consequential formation of isolated regions of Li metal, alongside a larger amount of decomposition reactions due to the porous morphology, will lower the Coulombic efficiency and hence shortening the cycle life.
Electrochemical performance of full cells
Ni-rich compounds are considered to be promising cathode materials for LIBs due to their large capacity, which in combination with a Li-metal anode results in a very high energy density19,51,52,53. In this context, NCM811 with a high active mass loading of 3.5 mAh cm−2 is used to test the performance of the prepared amide electrolyte. The high cathode mass loading aggravates the side reactions and requires a high efficiency of the Li-metal utilization54. Both Li||NCM811 cells, employing the different electrolytes, deliver similar initial charge and discharge voltage profiles, resulting in a specific capacity of 195 mAh g−1, shown in Fig. 6a, b. Upon subsequent cycling, the full cells with the amide electrolyte present much better cycling stability, retaining >88% of its initial capacity after 500 cycles as shown in Fig. 6c, and a high average Coulombic efficiency as shown in Supplementary Fig. 18. In comparison, the battery with the EC-based electrolyte rapidly decays to ~30% after 75 cycles as shown in Fig. 6a, c. Evaluation of the C-rate charging-discharging performance of the full cells, see Supplementary Fig. 19, demonstrates better rate capabilities for the amide-based electrolyte. This is in line with the smaller impedance observed, shown in Supplementary Fig. 20, which demonstrates the good compatibility of the amide electrolyte in full Li-metal cells. Furthermore, the full cycling history of a Li||NCM811 full cell with a restricted Li capacity is evaluated in Fig. 6c, where the Li-metal capacity on the Cu current exceeds the cathode capacity (~3.5 mAh cm−2) by a factor of 1.5. Under these demanding conditions stable cycling performance is achieved as >92% of the capacity is retained for 100 cycles. In addition, when FDMA is added to the conventional electrolyte system, as shown in Supplementary Fig. 21, the Li||NCM811 full cell using 1 M LiPF6-EC/DMC/FDMA (volume ratio 1:1:1) exhibits improved cycling stability compared with using 1 M LiPF6-EC/DMC.
Interface between cathode and electrolyte
The interface stability between cathode and electrolyte is further investigated by SEM and TEM. After 50 cycles in full cells, the morphology of the cathode materials is well preserved in the amide electrolyte, while obvious cracks and pulverization of the cathode are observed for the secondary particles in the EC-based electrolyte, see Supplementary Figs. 22 and 23. This indicates that the as-prepared amide electrolyte can suppress the structural degradation of NCM cathode materials. More evidence is provided by the transmission electron microscope (TEM) characterization. When cycled in the EC-based electrolyte, severe surface damage on the cycled cathodes is observed in Supplementary Fig. 24a, indicated by the yellow dashed lines. The corresponding fast Fourier transforms (FFT) patterns of the near-surface regions, up to 12 nm in depth, demonstrate the phase transition from the original layered structure to the rock-salt phase at the surface. In contrast, the surface layer on the particle cycled in the amide electrolyte is much thinner, ~4 nm, and more uniform. Moreover, it is mainly composed of amorphous cathode electrolyte interface (CEI) components as shown in Supplementary Fig. 24b. This indicates that this amide electrolyte effectively protects the cathode surface. Energy dispersive spectroscopy mapping clearly shows the presence of N- and F-containing components in the CEI layer, Supplementary Fig. 24c, explaining the decreased resistance and the improved electrochemical stability.
Furthermore, the chemical composition of CEI layers is investigated by XPS, results of which are shown in Supplementary Figs. 25–30. In the EC-based electrolyte, four peaks in the O 1s spectrum, located at about 530.5, 532.1, 533.4, and 534.3 eV, can be indexed as M–O (metal oxide from the cathode bulk), C=O, C–O, and Poly(CO3), respectively55. The appearance of the M–O species indicates that the CEI layer is cracked and non-uniform, exposing the cathode surface. In the O 1s and F 1s spectra of the amide electrolyte, a new peak indicates the presence of –CF3, which may originate from either the LiTFSI salt or FDMA. In addition, the CEI formed in the amide electrolyte is richer in LiF, as demonstrated by Supplementary Fig. 26. LiF has a large oxidative stability and excellent mechanical stability, which can be held responsible for the high-voltage stability of the CEI layer17,19. The N-containing species are also observed in the CEI layer of the amide electrolyte, which provides the required Li transport through the CEI layer. These results demonstrate that the amide electrolyte, through the synergy between both the FEC and FDMA, also provides a homogeneous and stable CEI protective layer, leading to the improved electrochemical performance of the full cells.