Fundamentals of aqueous chloride electrolytes
It is known that NH4+ in aqueous solution shares many similar properties with H2O,19,20 and can form hydrogen bonds with four H2O molecules.21 On the other hand, in dilute electrolytes, cations tend to be solvated by polar H2O molecules to form the hydrated cations.22 Based on these fundamental properties, our strategy involves introducing NH4Cl as a supporting salt in a dilute CdCl₂ solution to reconstruct the hydrogen bond network among H2O molecules and modify the aqua ligands of Cd2+ through the extra Lewis base of Cl- from NH4Cl, aiming to reduce water activity without compromising the original advantages of the dilute electrolyte. Starting with the baseline electrolyte of 1 M CdCl2 (referred to as 1M), we first examined the electrolyte structure of adding different concentrations of NH4Cl (1 M CdCl2 + 1 M NH4Cl, 1 M CdCl2 + 2 M NH4Cl, 1 M CdCl2 + 4 M NH4Cl, 1 M CdCl2 + 6 M NH4Cl, and 1 M CdCl2 + 7.5 M NH4Cl electrolytes referred to as 1M1M, 1M2M, 1M4M, 1M6M, and 1M7.5M, respectively). In Fig. 2a, Fourier transform infrared spectroscopy (FTIR) at the 1100 cm-1 band of NH3 was absent in the spectra of all NH4+-containing electrolytes. This indicated that the hydrolysis behavior of NH4+ to produce NH3·H2O and H+ can be considered negligible.23 In contrast, the peak of NH4+ appeared at 1440 cm-1 band and became increasingly prominent with higher concentrations of NH4Cl. They existed in aqueous solution as hydrated NH4+, which is confirmed by the presence of hydrogen bond between NH4+ and H2O, indicated by the bands at 3360, 3200, and 3050 cm-1 in the spectra.23,24 The impact of NH4+ on the hydrogen bond network of H2O molecules were analyzed through Raman spectra (Fig. 2b), wherein the O-H stretching for H2O molecules in the high-frequency range were deconvoluted into three peaks corresponding to the strongly, weakly, and slightly hydrogen-bonded components at the peaks of 3248.9, 3468.4, and 3628.8 cm-1 (Supplementary Fig. 1). From pure water to 1M7.5M electrolyte, the hydrogen networks of H2O molecules gradually transitioned from the strong hydrogen bond-dominated conformation to the weak hydrogen bond-dominated conformation; meanwhile, the peak intensity corresponding to NH4+ at the peaks of 2878.7, 3032.1, and 3115.0 cm-1 gradually increased (Fig. 2b and Supplementary Fig. 1). They demonstrated that the introduction of NH4+ into the 1M electrolyte disrupted the original hydrogen bond environment among H2O molecules and subsequently participated in forming a new hydrogen bond network between NH4+ and H2O as hydrate NH4+. Furthermore, as the concentration of NH4Cl increased, the shifts of 1H nuclear magnetic resonance (1H NMR) and 17O NMR signals of H2O also indicated the enhanced hydrogen bond interactions between H2O and NH4+ as hydrated NH4+ but decreased hydrogen bond interactions among H2O molecules (Supplementary Fig. 2). Besides the NH4+-involved hydrogen bond network, we observed that the formation of tetrachlorocomplex, [CdCl4]2−, was confirmed at the peak of about 260 cm−1 with the increased NH4Cl concentration, as shown in the low-frequency Raman spectra of Fig. 2c.25,26 This confirmed our design concept (Fig. 2d), where NH4Cl provided extra Lewis base of Cl− that complexed with Lewis acid of Cd2+ to transform the hydrated Cd2+ to [CdCl4]2−. Note that this transformation addressed a series of side reactions initiated by hydrated cations, such as hydrogen evolution reaction (HER) and dendritic growth, as discussed later. Therefore, the NH4Cl imparted versatility to dilute 1M electrolyte, involving rich NH4+-involved H-bond network and [CdCl4]2−.
Electrochemistry of Cd anodes
In our study aimed at developing practical energy storage infrastructures, we thus employed rigorous testing conditions to screen the tailored aqueous electrolytes. This involved evaluating the CE of Cd plating/stripping behaviors at a substantial AU of 33% (3 mAh cm−2), using a precise two-electrode Swagelok-type cell (Supplementary Figs. 3−5). Fig. 2e summarized the performance in different electrolyte. The average CE of Cd plating/stripping behaviors in 1M6M electrolyte was the highest and reached 99.89% (400 cycles), much higher than the average CE of 1M (99.76%) obtained from the reversible 129 cycles. It demonstrated the introduction of 6 M NH4Cl significantly improved the performance of the Cd anode. Consequently, the 1M6M electrolyte was identified as the optimal choice for subsequent research. We further utilized Cd//Cu cells to assess the performance of the ACB at a higher AU of 55% (5 mAh cm−2) and a higher current density of 10 mA cm−2, aiming to meet the demanding requirements of practical applications.27 The performance was compared against that of the baseline 1M electrolyte in the ACB and commonly used 2 M ZnSO4 (2ZS) electrolyte in the AZB,7 as depicted in Fig. 2f and Supplementary Fig. 6. An initial CE of 99.50% and an average CE of 99.93% over 300 cycles were achieved in 1M6M electrolyte, which reflect the highest Cd plating/stripping efficiency at a practical condition. By contrast, an initial CE of 98.89% and average CE of 99.86% within reversible 39 cycles were obtained in baseline 1M electrolyte, which are still obviously superior to the Zn plating/stripping efficiency of initial CE (94.20%) and average CE (98.51% within reversible 33 cycles). Furthermore, to figure out the highest reversibility of Cd anode in 1M6M electrolyte, we analyzed the anode morphologies after cycling by a scanning electron microscopy (SEM). It was observed that Cd anode in 1M6M electrolyte demonstrated dendrite-free capability with a smooth and dense morphology (Fig. 2g), thereby affording the excellently electrochemical performance. In contrast, the Cd anode in 1M electrolyte showed randomly oriented plate-like deposits (Supplementary Fig. 7a), whereas the moss-like dendrites that more easily leads to cell shorting formed on the Zn anode (Supplementary Fig. 7b).
Considering that energy storage infrastructures necessarily involve intermittent usage in practical applications, we further evaluated the aging CE of restoring aging-induced anode in aqueous electrolyte under an extremely stringent condition, as shown in Fig. 2h. We observed that long-term storage of metal Cd in aqueous electrolytes did not lead to capacity loss caused by corrosion. This was evidenced by stable voltage polarization curves of plating, aging and stripping processes (Fig. 2i), high average CE of ~99.34 (Fig. 2j), and highly overlapping plating/stripping curves over cycles (Fig. 2k). To our knowledge, no current metal battery system demonstrated this capability, even under conditions of minimal anode utilization. For example, the promising Zn anode only achieved an aging CE of less 75% (Fig. 2j) and suffered from obvious voltage polarization-induced cell failure in second cycle due to capacity loss caused by corrosion of aqueous electrolyte (Supplementary Fig. 8a, b).11 By the way, although Cd anode was cycled only 6 cycles in 1M electrolyte, its failure was not due to corrosion reaction but rather to cell shorting (Supplementary Fig. 8c, d). Additionally, it exhibited a high average aging CE of 97.58% (Fig. 2j). This further indicated that Cd anode is inherently corrosion-resistant in aqueous electrolytes.
Dendrite-free Cd plating/stripping behaviors
To further validate the dendrite-free capability of ACB and exclude homogeneous effect during plating process, we investigated Cd plating/stripping behaviors on Cu foil substrate at a high areal capacity of 5 mAh cm−2. The morphological evolutions were observed by ex-situ SEM (Supplementary Figs. 9−11). The Cd deposits in 1M electrolyte could maintain regular texture morphology up to 3 mAh cm−2, whereas, beyond this capacity, uncontrolled dendrites began to form and grow (Supplementary Fig. 9). In contrast, the plating process for Cd in 1M6M electrolyte consistently showed a regular grain growth process (Supplementary Fig. 10). However, the Zn plating process was accompanied by the pronounced moss-like dendrite growth (Supplementary Fig. 11).
We subsequently studied the cross-section morphologies of Cd deposits by a focused ion beam SEM (FIB-SEM) and their atomic arrangements using a high angle angular dark filed scanning transmission electron microscopy (HAADF-STEM), as shown in Fig. 3 a-f. Study areas were selected on the dendritic and smooth surfaces of Cd deposits (insets of Fig. 3a, d), then cross-sectioned perpendicular to the substrates (Supplementary Fig. 12). The polycrystalline structures with multi-boundaries were observed in Cd dendrite (1M), which indicated that the dendritic growth is accompanied by the formation new grain boundary and its subsequent grain growth (Fig. 3a). In addition, the HAADF-STEM images of adjacent areas of 1 and 2 (Fig. 3a) revealed discontinuous atomic arrangement (Fig. 3b, c). This denotes different growth directions of Cd grains, which is the root of dendrite formation and growth. In contrast, the large micron-size cross section of Cd deposit (1M6M) exhibited a single crystalline structure, as no grain boundary was detected (Fig. 3d). Furthermore, the HAADF-STEM image revealed a uniform hexagonal close-packed (HCP) atomic arrangement (Fig. 3e), on which Cd atoms can grow continuously and uniformly in a layer-by-layer manner, “abababab”, resulting in the densest atomic packing (Fig. 3f). Therefore, this grain growth pattern prevents the formation of new grain boundaries and, consequently, dendrite growth.
We subsequently quantified the charge-transfer kinetics in 1M and 1M6M electrolytes, by measuring the exchange current densities of Cd//Cd symmetric cells (Fig. 3g and Supplementary Fig. 13). Interestingly, the exchange current density in the 1M6M electrolyte is 3-fold higher than that the 1M electrolyte, reaching to 51.7 mA cm−2. Moreover,the 1M6M electrolyte exhibited an ultrahigh ionic conductivity of ~550 mS cm−1, researching a level comparable to that of H+/OH− conduction in aqueous medias28 and significantly superior to that of the 1M electrolyte (Fig. 3h). These results indicated the much faster kinetics of charge transfer and [CdCl4]2− desolvation to Cd2+ in 1M6M electrolyte than that of the charge transfer and the desolvation of hydrated Cd2+ to Cd2+ in 1M electrolyte (Fig. 3i). Therefore, this obvious difference in kinetics in 1M and 1M6M electrolytes contributed the inherently different Cd plating behaviors, one a kinetics-limited system and one a fast-kinetics system. Specially, the kinetics-limited 1M system could maintain the regular grain growth up to an areal capacity of 3 mAh cm−2, but resulted in the dendritic growth at a high areal capacity of 5 mAh cm−2 (Supplementary Fig. 9). It indicated that the relatively slow kinetics of charge transfer and desolvation cannot keep up with the rapid and continuous depletion of Cd2+ in the interfacial layer between electrode and electrolyte as the plated areal capacity increases, thereby disrupting the thermodynamic equilibrium necessary for complete Cd grain growth (Fig. 3i).29,30 Consequently, it became a kinetics-limited Cd growth process, readily inducing the formation of new grain boundaries and subsequent irregular grain growth.30,31 In contrast, in the 1M6M system, the ultrafast charge transfer and desolvation enabled the Cd growth process to proceed without kinetic limitations, resulting in a controlled and regular grain growth behavior (Fig. 3i). Therefore, it addressed the dendritic issue under deep cycling of battery.
Corrosion-resistant Cd anode
In addition to being free from dendritic issue, the corrosion-resistant capability of the metal anode is also paramount for long-term energy storage applications.22 Notably, the highest aging CE has highlighted the high corrosion-resistant capability of ACBs (Fig. 2j). To further deep study this characteristic, we employed 3 M H2SO4 for additional verification (Fig. 4a). Unlike metallic Zn, which underwent rapid chemical corrosion, gas evolution, and eventual complete dissolution in acidic solution, metallic Cd exhibited high resistance to chemical corrosion and retained its intact macro and micro morphologies. However, in a real aqueous environment, the corrosion scenario becomes significantly more complex due to the added factor of electrochemical corrosion.32,33 We found that since the redox potential of Cd2+/Cd is slightly lower than that of the HER, the Cd anode still faces the risk of electrochemical corrosion (Fig. 4b). We subsequently used the symmetric cells to magnify this risk at continuous current fluctuations of 0.5 mA cm−2 to induce HER, as shown in Fig. 4c.34,35 Given the markedly lower redox potential of Zn2+/Zn couple compared to HER (Fig. 4b), the Zn//Zn symmetric cell consistently experienced corrosion reactions and thus formed the byproduct of Zn4SO4(OH)6·H2O (Supplementary Fig. 14), manifesting in cell polarization and eventual failure after only 45 h (Fig. 4c). In contrast, the Cd//Cd symmetric cells exhibited superior corrosion-resistant capability in both 1M and 1M6M electrolytes, as evidenced by their stable and extremely low voltage fluctuations over 229 h (Fig. 4c). However, byproducts were still detected in 1M electrolyte system during amplified corrosion process, in contrast to smooth Cd anode in 1M6M electrolyte, appearing as fine crystalline Cd(OH)Cl particulates adhering to the Cd anode surface (Supplementary Fig. 15) due to the HER of hydrated Cd2+ to form Cd(OH)Cl and H2.
Considering that amorphous or minimal amounts of byproducts may not be detected by XRD, more quantitative analysis using X-ray photoelectron spectroscopy (XPS) revealed the depth-profiled composition throughout the Cd anode surface (Fig. 5d, e). Compared to the Cd anode in 1M, only superficial layers of bare Cd anode and Cd anode in 1M6M contained O 1s signals, which can be deconvoluted into Cd-OH and Cd-O bands (Fig. 5d). We attributed the presence of these signs to slight oxidation by air (Bare Cd and Cd in 1M6M) or very weak electrochemical corrosion (Cd in 1M6M). Therefore, these signs disappeared after Ar+ sputtering to the depths of 40 nm and 200 nm. However, notable O 1s signals corresponding to Cd-OH and Cd-O bonds were consistently detected on the surface of the Cd anode (1M) from a sputtering depth of 0 nm to 200 nm. Similar results were confirmed by the Cl 2p signals corresponding Cd-Cl bands (Fig. 5i), in which these notable signals were present on the Cd surface (1M) at various depths but only existed in the superficial layer of the Cd surface (1M6M). Consequently, our amplified corrosion experiments and the comprehensive characterizations demonstrated that the Cd anode in the 1M6M electrolyte is highly resistant to corrosion reactions. Given the variations in the local environment from 1M electrolyte to 1M6M electrolyte, we suggested that the replacement of H2O molecules by Cl− for hydrated Cd2+ to [CdCl4]2− significantly weakened the HER, thereby favoring the dominance of Cd2+ desolvation and subsequent deposition behaviors.
Electrochemical performance of ACBs
The electrochemical performance of ACBs was assessed using representative and commercial cathode materials, including coordination-type PANI, capacitance-type AC and intercalation-type V2O5, paired with the durable Cd anode. Cyclic voltammetry (CV) profiles were first collected to study the electrochemical behaviors of Cd//PANI full cell (Fig. 5a). The cell in the 1M6M electrolyte exhibited the relatively higher response current and reduction potential, indicating faster reaction kinetics and lower overpotential. Corresponding electrochemical impedance spectroscopy (EIS) demonstrated the smaller charge transfer impedance in the cell in 1M6M electrolyte (Supplementary Fig. 16). Furthermore, the rate performance of Cd//PANI full cells in 1M6M electrolyte showed a discharge capacity of 136 mAh g−1 at a low current density of 1 mA cm−2, which is lightly higher than the cell in 1M electrolyte (Fig. 5b, c). However, this performance difference widened as the current density increased. The cell in the 1M6M electrolyte still exhibited an applicable discharge capacity of 75 mAh g−1 at a high current density of 50 mA cm−2, whereas, in the 1M electrolyte, the discharge capacity of cell decreased to 54 mAh g−1. This suggested that the full cell in the 1M6M electrolyte exhibited faster charge transfer and reaction kinetics compared to the cell in the 1M electrolyte. This result was further confirmed by the high exchange current of full cell with the 1M6M electrolyte (Supplementary Fig. 17), indicating the capability of ACBs for high-rate performance.
Furthermore, long-term rechargeable capability of Cd//PANI full cells was evaluated at different high rates (denoted 130 mA g−1 as 1 C rate), as shown in Fig. 5d, e and Supplementary Fig. 18. It should be noted that although high rates are typically used in aqueous batteries to achieve extended cycle life, rechargeability at low rates remains significantly limited yet necessary.27,36 However, in our work, the full cell demonstrated its excellent low-rate rechargeable capability at 1.12 C, showing a high capacity retention rate of about 80% after 1,000 cycles (Fig. 5d). Furthermore, at a superhigh rate of 50.34 C, the full cell exhibited exceptional rechargeable capability, retaining 83% of its initial capacity after 20,000 cycles (Fig. 5e), which meets the high-power demands of large-scale energy storage applications. On the other hand, the ACBs demonstrated their high compatibility for multi-scenario application, as confirmed by pairing the Cd anode with capacitance-type AC and intercalation-type V2O5 cathodes. The Cd//AC capacitor showed typical characteristic of electronic double layer capacitors with large response currents at scan rates from 20 to 100 mV s−1 and performed excellent rechargeability with a capacity retention of about 90% upon 10,000 charging/discharging cycles (Supplementary Fig. 19). Attractively, the ACB with intercalation-type V2O5 cathode also performed stable cycling with a high discharge capacity of 270 mAh g−1 after 100 cycles (Supplementary Fig. 20).
From an application perspective, the use of a thin Cd anode with high utilization under deep cycling conditions is essential for practical energy storage infrastructures. Here, we further assembled the Cd//PANI full cell with a high-load cathode of 38.22 mg cm−2 and a low N/P ratio of 1.91 (anode capacity: 9.09 mAh cm−2). As shown in Supplementary Fig. 21a, the high-load full battery using 1M electrolyte could only run for 4 cycles before experiencing a sudden internal short circuit. In contrast, a significantly longer lifespan of 800 cycles was achieved in the full cell using 1M6M electrolyte, with a high cumulative areal capacity of 2.57 Ah cm-2 (Fig. 5f). Additionally, this high-load and low N/P full cell in 1M6M electrolyte, with a 12-hour rest period per cycle, exhibited a low self-discharge rate of only 5.4% after the first charge to 1.1 V, which further decreased to 2.4% after the twentieth charge to 1.1 V (Fig. 5g, h and Supplementary Fig. 21b). This demonstrated its capability for intermittent usage in practical applications.