Flexible liquid metal–polymer batteries assembled with a fluid GaInSn alloy anode, a flexible PANI cathode, and a GaCl3/NH4Cl/PVA hydrogel electrolyte and encapsulated in an elastic silicone elastomer shell were fabricated as 1-D (fiber), 2-D (sheet) and 3-D (sphere) structures, as shown in Fig. 1a and Supplementary Fig. 1. As illustrated in Fig. 1b–d, the battery pack consisting of 1-D and 3-D batteries connected in series could power a light-emitting diode (LED) either during the stretching and release of the 1-D battery or during various deformations and release of the 3-D battery.
2.1. Effect of Electrolyte Acidity on the Properties of Ga-Based Electrodes
Metal anodes commonly self-corrode due to the HER and acid attack in aqueous electrolytes, especially, with strong acidities. In the case of the liquid metal–PANI battery, although the use of a strongly acidic electrolyte can improve the activity and stability of the PANI cathode, it will aggravate the self-corrosion of the liquid metal anode, resulting in a low open circuit voltage and severe self-discharge.7, 52 Therefore, the degree of self-corrosion of Ga-based metal anodes in electrolytes with different pH levels was determined by Tafel extrapolation and linear scan voltammetry (LSV), and from the potential distribution during charging and discharging.
The corrosion potentials/rates of Ga, GaIn, and GaInSn as determined from the Tafel curves (Fig. 2a, Supplementary Fig. 2 and Supplementary Table 1) were − 0.87/0.126, -0.82/0.056, and − 0.79 V vs SCE /0.026 mmol per year (mmpy) in the pH 0.9 electrolyte, and − 0.76/0.080, -0.68/0.052, and − 0.65 V vs SCE /0.008 mmpy in the pH 1.6 electrolyte, respectively. The protection efficiencies of GaIn and GaInSn were 16.6% and 79.9% in the pH 0.9 electrolyte, and 31.6% and 94.9% in the pH 1.6 electrolyte, respectively. The result indicates that although the use of the pH 0.9 electrolyte led to more serious metal corrosion than the use of the pH 1.6 electrolyte, the presence of In and Sn in the GaInSn alloy significantly decreased the corrosion rate in each stage in both the electrolytes. The HER of Ga-based metals was measured to determine the nature of metal corrosion. As shown in Fig. 2b, the HER onset potentials of Ga, GaIn, and GaInSn are − 0.71, -0.93, and − 1.09 V vs SCE in the pH 0.9 electrolyte, and − 0.93, -1.05, and − 1.15 V vs SCE in the pH 1.6 electrolyte, respectively. The result shows that in both electrolytes (pH 0.9 and 1.6), the resistance to the HER increased in the order of Ga < GaIn < GaInSn, since the addition of Sn further increased the resistance of the alloy to the HER because of the high HER overpotentials of In and Sn. To evaluate the effect of electrolyte acidity on the chemical stability of Ga-based metal anodes in the batteries, symmetric cells employing Ga, GaIn, or GaInSn in pH 0.9 and 1.6 electrolytes were tested at a current density of 1.0 mA cm− 2 for 30 min. As shown in Fig. 2c, the overpotentials of the Ga-based electrodes were higher in the pH 0.9 electrolyte than in the pH 1.6 electrolyte because of the stronger acidity of the former; unlike the GaIn- and Ga-based cells that exhibited unstable charge–discharge voltage curves with large polarizations, the GaInSn-based cell exhibited a stable charge–discharge voltage curve over 150 h with a smaller voltage polarization in both the electrolytes. Therefore, among the Ga-based alloys examined in this study, GaInSn is the most suitable anode candidate for aqueous liquid metal batteries that use a strongly acidic electrolyte.
The cycling performance of the Ga–PANI batteries (Fig. 2d) revealed that the overall performance of the GaInSn-based battery was better in the pH 0.9 electrolyte than in the pH 1.6 electrolyte; the GaInSn-based battery with the pH 0.9 electrolyte exhibited a higher capacity of 252.6 mAh g− 1 after 500 cycles with a higher average Coulombic efficiency (CE) of 98.6% due to the higher ionic conductivity of the pH 0.9 electrolyte (Supplementary Fig. 3) and high activity of PANI in the strongly acidic electrolyte.53 In contrast, although the Ga- and GaIn-based batteries exhibited excellent capacities of 496.2 mAh g− 1 and 473.7 mAh g− 1 after the first few cycles in the pH 0.9 and pH 1.6 electrolytes, respectively, their capacities significantly decreased to below 50 mAh g− 1 after 150 cycles because of the low corrosion resistances of Ga and GaIn in the strongly acidic electrolytes, resulting in relatively low CEs (< 89%). Therefore, the GaInSn–PANI battery exhibited the best performance when GaInSn, the most stable alloy, was used as the anode. A comparison of the cycling performances of Ga–PANI, In–PANI, and Sn–PANI batteries (Supplementary Fig. 4) revealed that due to the chemical inertness of In and Sn, the In- and Sn-based batteries delivered considerably low capacities (< 15 mAh g− 1) during the first cycle, which then rapidly decreased to 5 mAh g− 1. This indicates that Sn and In are inactive for the battery reactions upon alloying with Ga. Thus, Ga is the only active constituent of the anode in the GaIn–PANI and GaInSn–PANI batteries.
To determine whether the liquid metal anode can stably operate in the acidic electrolytes, anodic and cathodic polarization potential curves were measured from the charge–discharge cycling test. As shown in Fig. 2e and Fig. 2f, after charging, the potentials of the GaIn anode were − 1.26 and − 0.98 V vs SCE, whereas those of the GaInSn anode were − 1.08 and − 1.01 V vs SCE in the pH 0.9 and 1.6 electrolytes, respectively. This, together with the LSV result, (Fig. 2b) indicates that the GaIn anode operates only in the pH 1.6 electrolyte. This is because in the pH 1.6 electrolyte, the anodic potential of GaIn (-0.98 V) is higher than the HER onset potential (-1.05 V), whereas in the pH 0.9 electrolyte, the anodic potential of GaIn (-1.26 V) is much lower than the HER onset potential (-0.93 V); this implies that the HER occurs in the pH 0.9 electrolyte, not in the pH 1.6 electrolyte. In contrast, GaInSn stably operated in both electrolytes as its anodic potential (-1.08 and − 1.01 V, respectively) is higher than the HER onset potentials in both. In the case of the pure Ga anode, a significant portion of the charging curve is below the HER onset potential for both electrolytes (Supplementary Fig. 5), which indicates that Ga is unsuitable as an anode in both electrolytes because of severe corrosion. Moreover, the PANI cathode exhibited a regularly fluctuating cathodic potential in the range of -0.2 to 0.6 V vs SCE during cycling, which avoids structural or morphological damage due to the over-oxidation and over-reduction of PANI.54, 55 This is essential for the stable performance of the GaInSn–PANI battery and is further confirmed by SEM analysis (Supplementary Fig. 6).
2.2. Working Mechanism of the GaInSn-PANI Battery
The working mechanism of the rechargeable GaInSn–PANI batteries was systematically investigated by scanning electron microscopy (SEM), energy dispersive X-ray spectroscopy (EDS), Raman spectroscopy, and X-ray photoelectron spectroscopy (XPS). As shown in Fig. 3a (Supplementary Fig. 7 and Supplementary Fig. 8), the Ga content of the GaInSn anode decreased from 67.68 wt% to 63.89 wt% after the 1st discharge, increased to 70.96 wt% after the 1st charge, and remained almost unchanged (70.48 wt%) after 100 charging cycles. To accurately evaluate the compositional changes in the anode, the mass fraction, i.e., Ga/Sn and In/Sn ratios (Sn being the most inert metal in the alloy) were determined; the results are shown in Fig. 3b. Because the activities of In and Sn are lower than that of Ga, In/Sn stabilized at approximately 2.07 over 100 cycles, which indicates that Sn and In are the inactive components of the liquid metal anode. In contrast, during charging–discharging, the Ga/Sn ratio (6.427) first decreased to 5.433 and then increased to 7.501, which indicates that Ga is the only active material in the GaInSn anode with the reaction Ga − 3e− \(\rightleftharpoons\) Ga3+.7, 8 Moreover, after 100 charging cycles, the Ga/Sn ratio remained almost constant (with a decrease of 0.52 wt% compared with that after the first charging cycle), which indicates the sufficient reversibility of the proposed anode even after 100 cycles for liquid metal secondary batteries. It is worth noting that the GaIn anode exhibits a similar working mechanism; however, some small solid In particles appeared on surface of the GaIn anode after discharging (Supplementary Fig. 9b and Supplementary Fig. 10) because the ratio of Ga and In in the GaIn anode exceed the liquid window during stripping (Supplementary Fig. 11). In contrast, even after long-term cycling test, the GaInSn anode still maintained the liquidity due to its wide liquid window (> 55wt% Ga maintain liquidity, shown in phase diagram in Supplementary Fig. 11).
Further, the Raman spectra (Supplementary Fig. 12) exhibited characteristic peaks at 1166.50 and 1480.17 cm− 1, which indicate the growth of PANI on the carbon substrate.56 In addition, the FTIR spectrum of the initial PANI cathode (Fig. 3c) shows peaks around 1493 and 1570 cm− 1 corresponding to the stretching vibrations of the quinoid diamine ring (N = Q = N) and benzenoid diamine ring (N–B–N), respectively, which imply that the synthesized PANI with the doped = N+– and un-doped = N– is partially oxidized.56, 57 And, during the 100 cycles, the N = Q = N and N–B–N signals show a chemical reversibility of PANI in the GaInSn–PANI battery. The XPS profiles of the initial PANI cathode (Fig. 3d) show peaks at 532.7, 399.0, 284.5, and 198.0 eV, which are assigned to O 1s, N 1s, C 1s, and Cl 2p, respectively. During discharge, the N = Q = N signal became more intense, while the N–B–N signal weakened, which is attributed to the reduction of PANI. In detail, after gaining an electron from the anode, the = N+– and = N– in PANI are reduced to –N– and –N−–, respectively. Meanwhile, the Cl− from PANI dissolves into the electrolyte, while Ga3+ is embedded into PANI via the reaction between Ga3+ and -N−-; this is supported by the disappearance of the Cl− peak and appearance of a Ga3+ peak in the XPS profile. In contrast, after charging, the N = Q = N signal weakened, while the N–B–N signal became intense, indicating the oxidation of PANI; this is confirmed by the reappearance of the Cl signal and disappearance of the Ga signal. The FTIR and XPS results indicate that Cl− and Ga3+, acting as counter-ions, participate in the redox reactions during the charging and discharging of the GaInSn–PANI battery, as in the case of previously studied Zn–PANI batteries.58, 59 The cation–anion hybrid mechanism of PANI cathodes has been proposed in previously reported PANI-based batteries with promising performances.56 In addition, the PANI cathode in the GaIn–PANI battery exhibited the same working mechanism (see the XPS profiles in Supplementary Fig. 13). Therefore, as shown in Fig. 3e, the reaction mechanism of the PANI cathode can be concluded as follows:
PANIoxd-(Cl)n + (n + 3m)e− +mGa3+ ⇌ (PANI)red-Gam + nCl− (1)
Furthermore, even after 100 cycles, the FTIR spectra remained unchanged and distinct Cl and Ga XPS signals appeared; this confirms the high stability of the PANI cathode in the GaInSn–PANI batteries. In general, during discharging, the Ga atoms that stripped from the GaInSn anode oxidize to form Ga3+ in the electrolyte, while the Cl in the PANI cathode reduces to form Cl−. In contrast, during charging, Cl− and Ga3+ return to PANI and GaInSn, respectively (Fig. 3f). In the charge–discharge process, the Ga from the GaInSn anode and the Cl associated with the PANI cathode act as reversible ions and participate in the redox reaction of the GaInSn–PANI battery. Thus, the overall reaction is as follows:
PANIoxd-(Cl)n + (1/3n + m) Ga ⇌3(PANI)red-Gam + 1/3n Ga3+ + nCl− (2)
2.3. Electrochemical Performance of the GaInSn-PANI Battery Under Various Temperatures and Deformability Degrees
The electrochemical performance of the GaInSn–PANI battery in the form of a 2-D sheet was evaluated. As can be seen from the long-term cycling performance (Fig. 4a) and galvanostatic charge/discharge profiles (Supplementary Fig. 14), the GaInSn–PANI battery delivered an initial discharge capacity of approximately 238.2 mAh g− 1 at 0.2 A g− 1 with a CE of 58.9%. The capacity increased significantly at the 2nd cycle (343.7 mAh g− 1), which is attributed to the activation of the partially oxidized PANI in the initial cycle. After 50 cycles, the capacity significantly decreased to approximately 223.9 mAh g− 1, because of the morphological and structural changes in the PANI cathode, and the irreversibly Ga3+ trapped in the CF@PANI electrode.60–62 After 500 cycles at 0.2 A g− 1, the GaInSn–PANI battery maintained a reversible capacity of approximately 179.8 mAh g− 1 with a capacity retention of 80.3% and an average capacity loss of 0.098 mAh g− 1 cycle− 1 (Fig. 4a, and Supplementary Table 2). Even at a high current density of 5.0 A g− 1, the GaInSn–PANI battery maintained a capacity of 92.8 mAh g− 1 with a high CE of over 98% and an average capacity loss of 0.013 mAh g− 1 cycle− 1 after 500 cycles. This indicates the promising electrochemical performance of the GaInSn–PANI battery at a high current density.
In addition to the outstanding cycling performance, the GaInSn–PANI battery exhibited an excellent rate performance. As shown in Fig. 4b, at low current densities of 0.2, 0.5, and 1.0 A g− 1, the battery delivered capacities of 212.1, 168.2, and 134.3 mAh g− 1, respectively. At even higher current densities of 2.0 and 5.0 A g− 1, the battery delivered capacities of 103.1 and 82.5 mAh g− 1, respectively. When the current density reverted to 0.2 A g− 1, the capacity recovered to 192.1 mAh g− 1, which is comparable to the initial capacity at 0.2 A g− 1. Moreover, after 500 cycles, the battery delivered capacities of 196.8, 171.4, and 132.4 mAh g− 1 at 0.5, 1.0, and 2.0 A g− 1, respectively (Supplementary Fig. 15). As can be seen from the Ragone plot (Fig. 4c), the GaInSn–PANI battery exhibits a significantly high energy density of 266.0 mWh g− 1 with a power density of 217.8 mW g− 1. At a high power density of 4300 mW g− 1, an energy density of 98.7 mWh g− 1 was achieved, which are higher than those of existing flexible metal–PANI batteries.50, 56, 63–66 Further, cyclic voltammetry (CV) was performed at different scan speeds (0.1–1.0 mV s− 1) to investigate the electrochemical behaviour of the GaInSn–PANI battery. Figure 4d shows that the area of the CV curves represents the capacity during the electrochemical process. The three peaks at approximately 0.79, 0.95, and 1.10 V indicate the different processes of Ga3+/Cl− reacting with the anode/cathode, and can be used to determine their corresponding capacitive effects according to the equation i = avb, where i (current) and v (scan rate) have a power law relationship.67 A b value close to 1 indicates a capacitance-controlled process, whereas a b value close to 0.5 indicates a diffusion-controlled process. The b values (slope) can be obtained from the log i vs log v plot. As shown in Fig. 4e, the b values of the three peaks are 0.81, 0.85, and 0.83, respectively, which indicates that capacitive contribution is the major contribution to total Ga3+ storage. The capacitive contribution is further confirmed by the equation i = k1v + k2v1/2, where k1v and k2v1/2 are the current contributions of the capacitance-controlled process and diffusion-controlled process, respectively. By quantification at every voltage and a certain scan rate, a simulated closed curve was obtained after integration (Fig. 4f). As can be seen, the capacitive process contributes to nearly 68.7% of the total Ga3+ storage at a scan rate of 0.6 mV s− 1; this suggests that the capacitive process dominates the entire reaction. This is the reason for the high-rate performance and excellent power density of the GaInSn–PANI battery.
Because of the liquid anode, the GaInSn–PANI batteries exhibited excellent cycling performance at low temperatures. As shown in Fig. 4g and Supplementary Fig. 16, the battery delivered capacities of approximately 168.2, 146.6, 103.4, and 64.0 mAh g− 1 at 25, 15, 5, and − 5°C, respectively, with over 95% CE. The decrease in cycling capacity with decreasing temperature is mainly attributed to the decrease in ion transfer in the poly(vinyl alcohol) (PVA) hydrogels.68 When the temperature was increased to 25°C, the capacity recovered to 148 mAh g− 1. This indicates the good reversibility of the GaInSn–PANI battery at various temperatures.
The mechanical and electrochemical properties of the GaInSn–PANI battery with a 2-D structure in stretched and bent states are shown in Fig. 5. When stretched to 200% (Fig. 5a, Supplementary Fig. 17 and Supplementary Video 1), the battery delivered a capacity (140.9 mAh g− 1) comparable to that in the pristine state (142.4 mAh g− 1). In addition, the battery exhibited a stable capacity during the bending (bending radius of 5 mm) and recovery processes. The Nyquist plots of the battery in various states are shown in Fig. 5b, and the simulated parameters obtained from the equivalent circuit of Re(RinQ1)(RctQ2)W (Supplementary Fig. 18) are summarized in Supplementary Table 3 and Supplementary Table 4. The starting point (Re) reflects the total contact resistance of the batteries, and the first and second arcs correspond to the interfacial resistance (Rin) and charge transfer resistance (Rct), respectively. Under the stretched and bent states, the Rin of the battery increased by 2–3.5 times and the Rct decreased by 33.7–39.2%, respectively, compared with those of the battery in the pristine state because of the increases in the interface contact area and charge transfer channels between the electrode and electrolyte. Under the combined effect of Rct and Rin, the battery exhibited a stable electrochemical performance during bending and stretching.
In addition to the battery-deformation degree mentioned above, the degree of battery-deformation difficulty will also be a necessary factor to be examined for wearable/deformable devices. To quantitatively exhibit the battery-deformation difficulty, the force information was recorded during stretching process. As shown in Fig. 5c, an average force of 0.850 N was required to stretch the battery encapsulated in a rubber shell to 200% of its original length, while almost the same force (0.842 N) was required to stretch the packing shell without the battery to 200% of its original length (Supplementary Fig. 19). This suggests that the battery body without packing requires a force of only several millinewtons (8 and 11 mN for 2-D and 1-D batteries, respectively, as shown in Supplementary Fig. 19 and Supplementary Fig. 20 and Supplementary Video 2) to stretch it to 200% of its original length. This is attributed to the outstanding fluidity of the GaInSn alloy at room temperature, and the special designed electrolyte and PANI cathode. Furthermore, the liquid metal battery demonstrated unique low-temperature flexibility: at -4.6°C, the battery could be easily stretched to 200% of its original length and recovered its original state upon release. The results indicate the excellent deformability of the GaInSn–PANI battery at a low temperature, and thus, its promising applicability to flexible devices.