Decoupling the origins of irreversible coulombic efficiency in anode-free lithium metal batteries

Anode-free lithium metal batteries are the most promising candidate to outperform lithium metal batteries due to higher energy density and reduced safety hazards with the absence of metallic lithium anode during initial cell fabrication. In general, researchers report capacity retention, reversible capacity, or rate capability of the cells to study the electrochemical performance of anode-free lithium metal batteries. However, evaluating the behavior of batteries from limited aspects may easily overlook other information hidden deep inside the meretricious results or even lead to misguided data interpretation. In this work, we present an integrated protocol combining different types of cell configuration to determine various sources of irreversible coulombic efficiency in anode-free lithium metal cells. The decrypted information from the protocol provides an insightful understanding of the behaviors of LMBs and AFLMBs, which promotes their development for practical applications.

2 cells. Thus, we have measured the proportion of dead Li and SEI in each cycle from one to five cycles of Li//Cu cells by the TGC method ( Supplementary Fig. 2). The results of the TGC measurements and electrochemical tests show that after the 1 st cycle, the irr-CEs of Li//Cu cell are quite comparable in the subsequent cycles, with the fraction of subsequent SEI gradually increased in each cycle. This can be explained as the instability of SEI and the subsequent fracture due to dendrite formation 2 . To conclude, since the proportion of dead Li and subsequent SEI needs to be quantified by TGC measurements, the two origins of irr-CE are combined in the proposed protocol and denoted as dead Li + sub. SEI. As for the difference between the 1 st cycle irr-CE and that of the 2 nd cycle, since that the irr-CEs after the 1 st cycle are quite similar to each other, indicating an additional irreversible reaction in the 1 st cycle, we attributed it to the 1 st cycle extra SEI formation due to the initial reductive electrolyte decomposition on the Cu surface when the cell is first discharged, namely 1 st extra SEI in the protocol.
Supplementary Fig. 2. TGC measurements quantifying the amount of dead Li 0 and SEI Li + associated with the total capacity loss of Li//Cu cells after one, two, three, four, and five cycles (the 1 st , 2 nd , 3 rd , 4 th , 5 th ). Supplementary Fig. 3. Charge and discharge curves of NMC//Li cell with deep discharge in the 2 nd cycle and normal discharge in the 1 st and 3 rd cycles, respectively.
As for NMC//Cu cells, we calculated the average CE after the transition state, namely in A/C ratio < 1 region.
, when n > t and A/C < 1.

Procedures to obtain irr-CEs in different cell configurations
In this section, the detailed step-by-step flowchart of the proposed protocol is provided as shown in Meanwhile, if the initial A/C ratio is larger than one, namely cathode is the limiting electrode, then the 1 st cycle irr-CE is dominated by cathode and equals to the fraction of the 1 st irr-capacity of cathode. In the subsequent cycles, the irr-CE can be separated into cathode degradation and sub. Ox. E.D. when A/C ratio remains larger than one, namely before the transition state. It should be noted that the fraction of cathode degradation in cathode//Cu cell could also be calculated from equation (1) like in cathode//Li cell; however, the value may not necessarily equal to that in cathode//Li cell considering the cathode degradation mechanism may be different among two cell configurations. However, as the active Li suffers from the continuous consumption by dead Li and subsequent SEI formation, the A/C ratio would eventually become less than one. Therefore, the anode electrode is then the limiting electrode and dominates the irr-CE of cathode//Cu cell, which is comparable to the aforementioned A/C < 1 case that the irr-CE can be separated into cross-talk effects and dead Li + sub. SEI.  Supplementary Fig. 5, there is also a discrepancy between the total irr-CE value and the sum of 1 st extra SEI formation and Li 0 amount from the TGC method.
This difference could originate from sub. SEI formation and the cross-talk effects in the 1 st cycle of the NMC//Cu cell.
As in the 2 nd and 10 th cycles, the increased discrepancy between the total irr-CE value and that of measured by the 5 TGC method can be again explained by the subsequent SEI formation and cross-talk effects. To conclude, the results from the validation are in consistent with those obtained from the integrated protocol, namely the major source of irr-CE in AFLMB should be attributed to the formation of dead Li due to poor reversibility of Li plating and stripping. Lastly, for further evaluation of the protocol under different circumstances, two example studies are conducted by changing the current density and electrolyte formulation in the following sections, respectively.

Supplementary Fig. 5 Inactive Li validation of Li//Cu and NMC//Cu cells for the irr-CE obtained from the proposed
protocol using the titration gas chromatography method.

Preparation of different amount of Lithium Metal
The

Comparison of Initial Overpotential and Polarization at 50% SOC
Supplementary Fig. 11 shows the corresponding initial overpotential, the polarization of Li//Cu, NMC//Li, and NMC//Cu cells under different parameters.
For a low current density of 0.2 mA cm -2 , Li//Cu cell shows larger initial overpotential and polarization at 50% SOC than Li//Li (Fig. S6a). This can be explained by the lower lithiophilicity of Cu, which leads to a larger Li nucleation barrier on Cu. It is also noticed that during the stripping process of Li//Cu cell, different from Li//Li cell, the second 8 plateau relating to the Li + diffusion from the bulk Li beneath the thick SEI formed due to the low reduction potential of metallic Li after cell assembly. This can be attributed to the substitution of Cu for Li, in which bulk Li does not exist thus the thick SEI is absent. In addition, it is found that the initial nucleation overpotential of Li//Li cell is strongly influenced by the length of resting time before cycling as shown in Fig. S1, the longer the cell rests, the higher the initial nucleation overpotential. As a result, when applying the proposed protocol, the resting time of each cell should also be considered and controlled.
Lastly, when using 5% of FEC as an additive in the commercial electrolyte, the initial overpotential and polarization at 50% SOC are both smaller than that in the electrolyte without 5% FEC except Li//Li cell, suggesting a more stable yet Li + conductive SEI formed after adding FEC into the electrolyte, which reduces the charge transfer resistance between the electrode/electrolyte interface. Li/Cu, using 1M LiPF6 in EC:DEC with 5% of FEC as electrolyte with the current density of 0.2 mA cm -2 . We selected NMC532 as the cathode material used in this electrolyte, one can substitute desired cathode material for cathode study in half-cell and AFLMB.

Extending the proposed protocol to ether system
We further extended our proposed protocol for evaluating the performance of AFLMB within the ether systems.
1M LiTFSI in DME:DOL (1:1) with 2 wt% of LiNO3 added and lithium iron phosphate (LiFePO4, LFP) are selected as the electrolyte and cathode material for demonstration, respectively. For the 1 st cycle irr-CE of Li//Cu cell ( Supplementary Fig. 14b), the total irr-CE of 1.91% can be separated into 0.31% of 1 st extra SEI formation and 1.60% of dead Li + sub. SEI. Subsequently, the irr-CE at the 2 nd cycle is 1.60% and keeps around ~1% in the subsequent cycles, which can be attributed to dead Li + sub. SEI. Meanwhile, LFP//Li cell shows the 1 st cycle irr-CE of 2.93% for 1 st cathode irreversible capacity. It is noted that the irr-CEs after the 1 st cycles are mostly greater than 1%, which is higher than those in carbonate systems and can be explained as greater amount of oxidative electrolyte decomposition in ether-based system than that in carbonate-based electrolyte due to the lower upper limit of potential window for ethers. Moreover, the fraction of cathode degradation in the irr-CE in the subsequent cycles can be calculated from the normalized capacity retention, which is ~0.17% in each cycle (see the gray bar in Supplementary Fig. 14b). Last but not the least, the 1 st cycle irr-CE in LFP//Cu cell is 11.30%, which is much higher than that of 1.91% in Li//Cu cell and 2.93% in LFP//Li cell. However, it is confirmed that the capacity loss in the 1 st cycle mainly originated from the loss of active Li inventory, since the 1 st cycle lost capacity of LFP could be restored after re-assembling the cycled LFP electrode with metallic Li anode (see Supplementary Fig. 15a). Based on the result, we suggest that the initial A/C ratio of LFP//Cu cell is less than one due to higher capacity loss at the anode and the irr-CE should be dominated by the anode. Thus, apart from the 0.31% of 1 st extra SEI formation and 1.60% of dead Li + sub. SEI, the remained 9.39% of irr-CE is suggested to be cross-talk effects. In the subsequent cycles, the irr-CEs in LFP//Cu cell are still larger than those in Li//Cu cells, suggesting cross-talk effects hold a remarkable proportion of irr-CE in LFP//Cu cell when using 1M LiTFSI in DME:DOL (1:1) with 2 wt% LiNO3 added as the electrolyte. Thus, the irr-CE in the 2 nd cycle can be attributed to 1.6% dead Li + sub. SEI and 2.56% cross-talk effects, and that in the 20 th cycle to 1.06% dead Li + sub.
SEI and 1.01% cross-talk effects. In addition, we reassembled the cycled LFP electrode (from the dead LFP//Cu cell) with metallic Li anode to confirm the origin of capacity loss in LFP//Cu cell ( Supplementary Fig. 15b). The results show that the reversible capacity of LFP//Li cell is regained back to the level of the 1 st charge capacity, thus, suggesting the capacity loss in LFP//Cu cell is mainly due to the irreversible reactions from anode, namely dead Li and SEI formation and the cross-talk effects. From the information obtained from the integrated protocol, it can be concluded that dead Li formation is significantly suppressed in ether-based electrolyte with the addition of LiNO3, which is similar to the previous works reporting the stabilizing effect of LiNO3 to metallic Li anode 3 . Thus, the capacity decay rate of the anodefree LFP//Cu cell is remarkably retarded compared to that of NMC//Cu cells in carbonate systems.