4-Electrode Full Cells for Operando Li⁺ Activity Measurements and Prevention of Li Deposition in Li-Ion Cells

Reference electrodes have been used by several authors in different configurations.^{1,15,39,40,47} The effect of the position of the reference electrode on the measured value can directly be seen in measurements with two reference electrodes in a 4-electrode cell—one between anode and cathode (Ref1 in Fig. 3) and another behind the cathode (Ref2 in Fig. 3).

Figure 4a shows that the voltage curves during charging of 4-electrode full cells for the 1st, 2nd and 5th cycle are similar. Indeed, we found that the reference electrode Ref1 was still largely intact after cycling (see inset of Fig. 4a). However, weak pitting corrosion and ongoing SEI formation on the reference electrode surface in the dynamic environment between anode and cathode might be the cause for the slight changes in the curves of the different cycles in Figs. 4a, 4b.

Zoom In Zoom Out Reset image size

The voltage difference between the reference electrodes (E_Ref1-Ref2) is shown in Fig. 4b. It can be seen that E_Ref1-Ref2 changes during charging. Figure 4c depicts the derivative dE_Ref1-Ref2/dQ of the data from Fig. 4b. Again, the data for the first few cycles show very similar trends. These measurements show that the position of the reference electrode is very important. For example, in aqueous lab cell setups, a position of the reference electrode near the working electrode is typically achieved by using a Luggin-Haber capillary.⁴⁸ For Li-ion cells, a position between anode and cathode is also supported by simulations by Dees et al.⁴⁹ and Delacourt et al.⁵⁰

Besides measurements of the anode potential, 4-electrode pouch full cells are also helpful to measure the Li⁺ ion activity a[Li⁺] in front of the anode during charging. By measurement of changes of a[Li⁺] it is possible to investigate the influence on the Li⁺ insertion reaction into the graphite particles, the kinetic properties of the Li⁺ insertion reactions, and their influence among each other. Since Li⁺ ions are present in solution, Li metal electrodes are essential for such measurements, since they are electrodes of the first kind.

The measurement principle of the 4-electrode pouch full cell is the a[Li⁺] dependency of a metallic Li electrode. The Nernst equation describes this dependency:

$$\begin{eqnarray}&&{E}_{Ref}={E}^{0}+\displaystyle \frac{RT}{zF}\cdot \,\mathrm{ln}\left(a\left[L{i}^{+}\right]\right)\end{eqnarray} \tag{ 1 }$$

In a 4-electrode setup (Fig. 3), it is possible to measure changes in the reference potential by measuring the voltage E_Ref1-Ref2 between the Li reference electrode between anode/cathode (Ref1) and the Li reference electrode in the inactive area behind the cathode (Ref2).

It has to be noted that the Nernst equation is a thermodynamic principle and therefore it is only valid in equilibrium situations. However, firstly in our setup there is no current flow between the two reference electrodes. Secondly, in our Li⁺ activity measurements, we apply only low C-rates in the order of 0.2C to stay near equilibrium conditions.

In the following we assume that a[Li⁺]_Ref2 = const. in the electrochemically non active area of Ref2:

$$\begin{eqnarray}\begin{array}{lll}{E}_{Ref1-Ref2} & = & {E}_{Ref1}-{E}_{Ref2}\\ & = & \displaystyle \frac{RT}{zF}\cdot \left[\mathrm{ln}\left(a{\left[L{i}^{+}\right]}_{Ref1}\right)-\,\mathrm{ln}\left(a{\left[L{i}^{+}\right]}_{Ref2}\right)\right]\end{array}\end{eqnarray} \tag{ 2 }$$

The fact that the potentials of both reference electrodes are present in this equation, underlines the importance of two reference electrodes to measure a[Li⁺]. Due to the connection in the experimental setup shown in Fig. 3, an increasing a[Li⁺] in the electrolyte between anode/cathode results in a more negative voltage difference E_Ref1-Ref2. As Fig. 4b shows, E_Ref1-Ref2 is negative but changes during charging. This means that a[Li⁺] in front of the anode surface does also change during charging, although the current is constant.

Additionally to the anode potential, a[Li⁺] can be used as a criterion for Li deposition.⁴⁶ Danner et al. found in simulations that the conditions for Li deposition expands from the anode surface to the current collector during charging.⁴⁶ For thick anodes, the authors reported that transport limitations can lead to strong Li⁺ concentration gradients (and consequently a[Li⁺] gradients) into the depth of anodes.⁴⁶ Since Li deposition can only happen where Li⁺ ions are present, Danner et al. did not find Li deposition near the current collector.⁴⁶ Similarly, it is likely that if a certain value of a[Li⁺] is exceeded in front of the anode surface, Li deposition will happen due to super saturation.⁴³

Figure 5a shows that the anode potential vs Li/Li⁺ for a graphite/Li half cell charged a 0.2C at 10 °C decreases steadily (black curve). At this low C-rate, different graphite phases can be seen from this curve. For comparison the dE_Ref1-Ref2/dQ curve is also plotted in Fig. 5a (red data points). The dE_Ref1-Ref2/dQ curve does not change steadily, however, it shows different peaks corresponding to a[Li⁺] changes. Interestingly, these a[Li⁺] changes occur at the transitions of graphite stages.

Zoom In Zoom Out Reset image size

Figure 5b shows a plot of dQ/dE_Anode derived from Fig. 5a together with the logarithm of the Li⁺ solid state diffusion coefficient in graphite taken from Funabiki et al.⁵¹ The data show that the diffusion constant is lowered at the graphite staging transitions. This means that the solid-state diffusion of Li⁺ in graphite is hindered several times during CC charging. Since the charging current during CC-charging is constant, a similar number of Li⁺ ions will arrive at the anode, independent from the diffusion inside the graphite. Consequently, a[Li⁺] is rising in the electrolyte on the anode surface, which is in accordance with the dE_Ref1-Ref2/dQ data from the 4-electrode full cell. This is illustrated by a simplified model in Figs. 6a and 6b for slow and fast diffusion of Li in graphite at CC charging, respectively.

Zoom In Zoom Out Reset image size

Furthermore, this coincides with the anode potential lowering stepwise at the phase transitions of lithiated graphite, since this increases the probability for Li deposition. Due to the constant current, it will be more likely that a larger part of the Li⁺ ions are deposited on the anode surface as a parallel reaction to intercalation.

This is consistent with the trends favoring Li deposition: lower temperatures,^{1,3,15,25,30,41} higher SOCs,^15,42 and higher charging C-rates.^3,15 While lower temperatures lower the probability to cross energy barriers, higher SOCs lead to slower solid diffusion in graphite by increasing the barrier heights. Similarly, higher charging currents lead to higher a[Li⁺] values near the anode in cases where the transport of Li⁺ ions to the anode is faster than intercalation.

This brings the cathode kinetics into play. Figure 7a shows the changes in the derivation of the reference electrode Ref1 in a NCM/Li half cell at 0.2C/10 °C. Figure 7b shows the solid diffusion coefficient of Li⁺ in NCM111 redrawn from Shaju et al.⁵² Obviously, the diffusion constant in NCM111 as a function of lithiation state (Fig. 7b) is different from that of graphite (Fig. 5b). However, both electrodes are involved in the charging process in a full cell (de-lithiation of cathode/lithiation of anode) and therefore there is an interaction between both.

Zoom In Zoom Out Reset image size

Our measurements showed that this significant interaction between anode and cathode is not negligible. Therefore, measurements of anode potentials should be carried out in full cells (e.g. graphite/NCM) rather than in half cells (e.g. graphite/Li).

Ideally, the Li⁺ transport inside the cathode, which includes Li⁺ diffusion in the active material and in the electrolyte inside the electrode (tortuosity respectively mass loading and density) should match that of the anode during charging, e.g. by a tailored blend cathode which has a similar Li transport properties as a function of SOC like the anode.

This could avoid an Li⁺ ion accumulation in front of the anode surface, which was already described Danner et al. in a model based analysis.⁴⁶ For non-CC charging, this could lead to an improved fast-charging capability. In the next section, we show that this can also be achieved partially by optimized charging protocols in which the current is reduced during charging.

Charging protocols can significantly change cycling aging behavior.^15,53–55 One important category of charging protocols is based on a stepwise reduction of the charging C-rate, either with constant current (CC) or constant voltage (CV) steps. For such types of charging protocols, we suggest abbreviating n CC steps as (CC)ⁿ, e.g. (CC)² for a 2-step protocol. Similarly, standard CC-CV charging would be named (CC-CV)¹ = (CC-CV).

For example, based on simulations, Remmlinger et al. suggested a charging procedure in which the anode potential remains positive by reducing the charging current.⁵⁶ Since the charging current is continuously reduced, this corresponds to a (CC)ⁿ protocol where n is a large number. Another approach was suggested in the patent by Bhardwaj et al., who used a 3-step (CC-CV)³ charging protocol in 3-electrode full cells, however, the CV-phases are comparably long.⁴⁰ We recently applied a 2-step (CC)² protocol to 3-electrode pouch full cells.⁵⁴ In that study, the 3-electrode full cells with reference electrode were reconstructed from graphite anodes and NCA cathode of commercial 3.25 Ah 18650-type cells.⁵⁴ Comparing long-term cycling tests with the original 18650 cells with optimized (CC)² and standard CC-CV charging protocols resulted in significantly enhanced cycle life for the tests without Li deposition.⁵⁴ However, the charging time was not enhanced compared to the conventional CC-CV protocol.⁵⁴

Figure 8 shows an optimized (CC-CV)⁴ charging protocol. In this charging strategy, the Li⁺ super-saturation on the anode surface caused by too high charging currents is prevented. The reduced current (green curve lower part of Fig. 8a) during the CV step leads to an decrease of a[Li⁺] (blue data points in lower part of Fig. 8a) and the anode potential vs Li/Li⁺ does not drop below 0 V vs Li/Li⁺ (red curve in upper part of Fig. 8a).

Zoom In Zoom Out Reset image size

Figure 8b shows a comparison of different charging protocols. For (CC) charging at 0.5C (black curve in Fig. 8b) the cell can be charged within 2 h, however this leads to negative anode potentials vs Li/Li⁺ and consequently to Li deposition (black dotted line). A reduction of the charging rate 0.2C leads to positive anode potentials (blue curves), avoiding Li deposition, however, the charging time is increased to ∼5 h. In contrast, the (CC-CV)⁴ charging protocol (red curves) allows to charge the cell in the fastest way (∼1.7 h) while retaining positive anode potentials vs Li/Li⁺. Compared to the 0.2C protocol the charging time is decreased by ∼68%.

Recently, we had reconstructed graphite anodes and NCM cathodes from commercial 16 Ah pouch cells into 3-electrode full cells with a Li reference electrode.¹⁵ These cells were systematically measured with different combinations of charging C-rates, ambient temperatures, and until various end-of-charge voltages.¹⁵ The results of the anode potentials in our lab were in accordance with long-term cycling tests with the original 16 Ah pouch cells performed at other EU-based institutes, i.e. faster aging correlated with for lower anode potentials.¹⁵ In the following, we show that the data from that paper¹⁵ can be used to construct an optimized (CC)⁵ charging protocol for the 16 Ah cells which is avoiding negative anode potentials.

Figure 9a shows the current as a function of charging time for standard (CC-CV) charging at 3C (blue dashed curve) and optimized (CC)⁵ charging (black solid curve). Please note that the starting current is 48 A, which could in principle be higher, however in this example it was limited by the equipment (max. 50 A). Furthermore, please note that for both charging protocols, the same amount of charge (12 Ah) is stored in the cells for the first cycle (both protocols charge from 10% to 90% SOC).

Zoom In Zoom Out Reset image size

The data in Fig. 9 is exemplarily shown for an ambient temperature of 25 °C. For other ambient temperatures different parameters must be used. We further note that the heating of the 16 Ah cells due to current flow was not taken into account here. From previous tests it is known that the smaller 3-electrode pouch full cells show nearly no heating compared to larger cells.^54,57 The reason is the lower number of electrode layers and the different volume/surface ratio of the cells, leading to a stronger heat dissipation of the smaller cells.⁵⁷

For CC-CV charging, the CV-phase is long compared to the CC-phase (Fig. 9a). Therefore, 37 min are needed to charge the cell. In contrast, the (CC)⁵ charging protocol needs only 23.4 min to charge the same capacity into the cell, corresponding to a reduction by ∼36% charging time, while 12 Ah are charged in both cases.

In CC-CV charging, 4.2 V is reached quite early which is then kept constant and consequently the current is decreased. In contrast, the current is kept constant until a certain critical voltage is reached in the (CC)⁵ protocol. These critical voltages correspond to transitions to negative anode potentials measured in 3-electrode pouch full cells.¹⁵

Comparing the two voltage profiles in Fig. 9a one can see that the current decrease in both charging protocols is similar but not the same. The CC phases in the (CC)⁵ protocol are in most cases longer compared to the CV-phase of the (CC-CV) protocol. This leads to more charge stored earlier in the (CC)⁵ protocol. This also leads to the faster charging and can be seen in the area of the (CC)⁵ current profile over the CV curve, since this area corresponds to a charge current vs time diagram.

Figure 9b shows a comparison of the two different charging protocols in long-term cycling tests with 16 Ah pouch cells. The capacity retention is based on a comparison of the ratio of the capacity in each cycle with the capacity of the first cycle. It can clearly be seen that the cycle life for the (CC)⁵ protocol is significantly enhanced. While the cell with the standard CC-CV protocol shows a capacity retention of ∼81% after 4000 cycles, the cell with the (CC)⁵ retained ∼89% after the same number of cycles. The improved charging strategy leads to an enhancement of cycle life of in the order of ∼170% with both protocols tested with the commercial 16 Ah cell. This improvement in the cycle stability is in accordance with our earlier results on the (CC)² charging with 3.25 Ah 18650-type cells, which showed an improvement of ∼200%.⁵⁴ However, for the charging protocol presented here, the charging time was reduced, which was not the case in our earlier study.⁵⁴