Introduction

Despite its simple structure, water is the most important yet least understood liquid on earth1,2,3,4,5,6,7. The transition from liquid water to solid ice is one of the most common, but complex, whereas vital phenomenon in nature. Undesired ice accumulation can be harmful to human survival and has been linked to aircraft crashes8, traffic accidents9, tissue/organ inactivation10, and battery failure11,12,13,14. Various strategies for preventing ice formation, including vapor heating, gelation treatment and use of hybrid solvents, have been extensively developed15,16,17,18,19. However, these methods are costly, inefficient and can even give rise to environmental pollution. Moreover, the underlying mechanism and driving force for ice nucleation and growth processes remain uncertain20,21. The design of an effective antifreeze technique requires a full understanding of the microscopic behavior of water molecules in order to realize practical application, notably the use of antifreezing electrolytes in batteries.

The introduction of various salts into aqueous solution is a simple and low-cost antifreeze strategy. Strong interaction between ions disassociated from dissolved salt solutes and the surrounding water can disrupt the hydrogen bond (HB) networks between water molecules. In addition, the electric field associated with the hydrated ions contribute to the rearrangement of dipolar water molecules. The role of different ions in impacting on the structure of water molecules is the subject of ongoing debate22,23,24,25,26,27,28. It has been demonstrated that several counterions at the polyelectrolyte brush interface can effectively regulate the dynamics of interfacial water and influence the ice nucleation processes29. The specific effects of ions on water structure have been classified into “structure-making” and “structure-breaking”, which is also applied to explain a variety of phenomena in electrolyte systems30,31. However, to the best of our knowledge, the thermodynamic origin of how ion specificity determines the freezing point (Tf) of water has not been fully understood so far. In the case of batteries operated at ultralow temperatures, antifreezing capability is not the only determining factor; dynamic ion migration of the aqueous electrolyte must be factored in, which is also related to ion-water interaction.

In this study, we utilize tetrahedral entropy32,33 (SQtet, quantifying the distribution of the local tetrahedral order) to bridge the interconnection between salt solutions with different anions and the corresponding antifreeze behavior. Through in-depth studies of the HB structure, kinetic and thermodynamic properties using experimental and theoretical methods, we find that the SQtet of water is vital in determining the significant difference in Tf of various salt solutions. The magnitude of SQtet is related to both kinetic and thermodynamic properties of the water molecules. Kinetically, faster dynamic behavior such as a higher self-diffusion coefficient and shorter orientation relaxation time facilitates more microscopic states, resulting in a higher SQtet within a certain time scale. In terms of thermodynamics, water molecules with higher SQtet exhibit lower Tf, which can be achieved by tailoring “structure-breaking” ions. As a proof of concept, the cation is fixed and four anions including SO42-, Cl-, Br- and ClO4- are systematically investigated. We find that the “structure-breaking” anion ClO4- endows water molecules with the highest SQtet, and contributes to the best frost resistance. In contrast, “the structure-making” anion SO42-delivers the poorest antifreeze ability, originating from its lowest SQtet of water. Moreover, we correlate Tf with the tetrahedral entropy in different aqueous salt solutions, confirming the crucial role of SQtet in determining water freezing behavior. As validation, we successfully fabricate polyaniline (PANI)||Zn aqueous zinc-ion batteries (ZIBs) utilizing the 5 m Zn(ClO4)2 electrolyte with a high ionic conductivity (1.12 mS cm−1) at −80 °C, which demonstrate enhanced antifreezing performance and stability. A high capacity (74.17 mAh g−1) is achieved at 1 A g−1 with ~85% capacity retention over 1200 cycles at −80 °C. In addition, 100% capacity is retained after 5000 long-cycles at −70 °C, which can satisfy global extreme cold regions. We also introduce a triboelectric nanogenerator (TENG) to successfully charge the fabricated ZIBs at −80 °C, providing potential application scenario as a self-charging power pack in extreme environments.

Results

Guidelines for antifreezing electrolyte

Thermodynamically, the liquidus temperature of electrolytes depends on the Gibbs free energy of solid and liquid (Fig. 1a). In order to lower the Tf of the electrolyte, a decrease in the Gibbs free energy is an essential strategy. According to the thermodynamic equation below,

$$G=H-{TS}$$
(1)

where G (kJ mol−1), H (kJ mol−1), T (K) and S (J mol−1 K−1) represent the Gibbs free energy, enthalpy, temperature and entropy of the actual system, respectively. The decrement of Gibbs free energy can be achieved by increasing the entropy of the solvent molecules. Thus, tailoring salty solution with high-entropy water is an effective approach for the design of an antifreezing electrolyte. We selected four salts with different anions (ZnSO4, ZnCl2, ZnBr2 and Zn(ClO4)2) representing a Hofmeister series for subsequent study (Fig. 1b). The “Hofmeister series” was first proposed to determine the ability of different salts (considering both type and concentration) to stabilize proteins in water. This arrangement is also termed lyotropic sequences or ion specificity series, and has been accepted widely. According to the definition of the series, ions can be classified as structure makers (kosmotropes) and chaos makers (or structure breakers, chaotropes). The former possesses a greater ability to retain order structure, while the latter disrupts the original structure. As shown in Fig. S1a, each water molecule can form four HBs due to the two lone pair electrons on the O atom. Consequently, abundant tetrahedrally structured water molecules occur in pure water. The introduction of a “structure-breaking” salt usually results in a breakdown of the original HB network between water molecules through dipole-dipole interaction (Fig. 1c, left), resulting in a faster diffusion dynamics and higher entropy (S). Some water molecules form a solvation shell around ions, restricting movement to some extent with a consequent decrease in S (Fig. 1c, middle). Therefore, in salt solutions, water molecules binding to ions or to other water molecules compete in a dynamic chemical exchange process (Fig. 1c, right). Both features should be taken into consideration simultaneously when tailoring the entropy of water.

Fig. 1: Schematic illustrating the design concept for a low-Tf electrolyte.
figure 1

a Relationship between Gibbs free energy, entropy of liquid and the freezing point, in which S represents entropy of systems. b Hofmeister series with four different anions (top) and the models of the corresponding four aqueous zinc salt solution for MD simulations (bottom). c The original water molecules form a tetrahedral network structure through hydrogen bonding. When the network composed of tetrahedrally structured water is disturbed by the introduction of ions, the water diffusion dynamics and S increase simultaneously (left). In addition, a solvation shell forms around the ions. As a result of the restriction imposed by the ionic electric field, the diffusion dynamics of water molecules is slowed with a reduction in S (middle). Water molecule bonding to an anion and other water molecules serve as two competing dynamic chemical exchange processes (right).

Full size image

Solvation and water structure analysis

As noted above, specific anions impose different impacts on the structure of water in aqueous solution, thus modulating the S and determining Tf. In order to corroborate our proposed mechanism, classical molecular dynamics (MD) simulations and experimental spectra were subjected to a combined analysis. The average HB number statistics derived from MD simulations for four electrolytes were first performed to assess differences in water structure, where the resultant order (ClO4-<Br-<Cl-<SO42-) matches the Hofmeister series, as indicated in Fig. 2a. A continuous HB network is difficult to form around the ClO4- anion due to its “structure-breaking” characteristics. Fourier transform infrared (FTIR) and Raman spectra were obtained to further confirm the HB statistical results. After background subtraction, FTIR bands (400~600 cm−1, ωB1, O:H-O bending vibration mode34) and Raman spectra (2800~4000 cm−1, ωH, O-H stretching vibration mode34) for pure water and the four electrolytes can be normalized with respect to peak area, as shown in Fig. S2a, b. The normalized spectra of the four electrolytes with a subtraction of the pure water component yields differential phonon spectra (DPS). The addition of the four salts all resulted in a red shift of ωB1 and blue shift of ωH with respect to pure water (the peaks and valleys correspond to phonon abundance gain and loss, respectively). The observed trends indicate that ionic electric-polarization can effectively break the intrinsic HB network, soften the O:H phonon, and stiffen the O-H phonon (Fig. 2b–d). It is found that ΔωH decreases in the order ClO4->Br->Cl->SO42-, establishing the most pronounced effect for ClO4- on HB perturbation (Fig. 2c).

Fig. 2: Theoretical and experimental analysis of the HB structure and dynamic relaxation behavior of water.
figure 2

a The average HB number of water clusters in four 5 m ZnX (X = ClO4-, Br-, Cl- and SO42-) electrolytes collected from MD simulations. Comparison of the DPS of the four electrolytes with respect to the reference spectrum of pure water, which derived from (b) FTIR and (c) Raman spectra. d Schematic showing HB breakage for two water molecules, leading to an elongation of the O:H non-bond and shortening of the O-H polar bond, representing a softening of the O:H phonon and stiffening of the O-H phonon. The proportion of water with strong (e) and non-HB (f) in four electrolytes at different concentrations fitted from Raman spectra. g The simulated MSD versus time curves for water molecules in four electrolytes. 2D LF-NMR T1-T2 relaxation spectra of (h) ZnSO4 and (i) Zn(ClO4)2 electrolytes. j Schematic showing water structure around ClO4-, Br-, Cl- and SO42-. Specified are the time scales associated with orientation relaxation τOR of water molecules in the first hydration shell.

Full size image

To extract more detailed information on HB structure and distribution in different electrolytes, the O-H stretching vibration mode of water was deconvoluted into three Gaussian components:19 strong HB (~3230 cm−1), weak HB (~3450 cm−1) and non-HB (~3620 cm−1), as shown in Fig. S3. With an increase in ClO4- concentration, the broad peak shifts to a higher frequency with a narrower shape, suggesting that the ratio of strong HB decreases while non-HB increases (Figs. 2e, f and S4a). In contrast, the degree of peak shift with a variation of SO42- electrolyte concentration is relatively minor, implying a negligible enhancement in non-HB (Figs. 2f and S4d). Based on the quantitative comparison of water in different environments (Fig. S4), the HB destruction capability of four anions follows the order ClO4->Br->Cl->SO42-, in accordance with the MD simulation results. Moreover, the self-diffusion coefficient of water derived from mean square displacement (MSD) curves also follows the sequence ClO4->Br->Cl->SO42- (Figs. 2g and S5), which is in good agreement with the data obtained from 1H diffusion ordered spectroscopy (DOSY); see Fig. S6. The consistency of these properties can be explained by the more severe disruption of the HB network releasing more free water molecules and facilitating faster self-diffusion.

The 2D low-field nuclear magnetic resonance spectra (LF-NMR) provide relaxation dynamics information for water molecules. The time regions (T2, spin-spin relaxation time) in the range 1-10 ms, ~100 ms and ~1000 ms were assigned to bound water, immobilized water and free water, respectively35,36. The free water T1(spin-lattice relaxation time) was almost equal to T2 and located on the diagonal of the 2D T1-T2 spectra (Fig. S7a). As demonstrated in Figs. 2h, and S7b, c, excess free water is found in SO42, Cl- and Br- related electrolytes while the proportion of immobilized and bound water can be neglected, which is similar to that observed for pure water. In contrast, appreciable immobilized and bound water is found in the ClO4--based electrolyte (Fig. 2i). Therefore, the HB number between water molecules can be significantly reduced. The HB number between water and the ClO4- anion is greater than that of SO42-, which can be further confirmed by the HB analysis presented in Fig. S9, and the FTIR measurements in Fig. S10 (23.14 cm−1 redshift for ClO4- and 16.39 cm−1 redshift for SO42-). It should be noted that the T2 of free water in ClO4--based electrolyte is larger than that of SO42-, suggesting a faster water molecule relaxation process, consistent with the MSD simulation results. We have also investigated the time scales associated with orientational relaxation (τOR) in the first hydration shells (1HS) of four anions (Figs. 2j and S11). In the case of ClO4-, the orientational relaxation of water occurs in ~ 6.15 ps, shorter than that of Cl- and Br- (~23.14 ps and 55.19 ps, respectively). Due to the higher valence of SO42-, the water molecules are tightly bound and more sluggish (τOR~2.22 ns). Hence, the 1HS of ClO4- is lapsable and the inner bound water are freer to rotate, contributing to a higher rotational entropy than the other anions.

To resolve the local structure and diffusion barrier of water molecules in our models involving different anions, radical distribution functions (RDF) and the potential of mean force (PMF) were employed. In Fig. S12a, the RDF of SO42--H (H2O) shows a well resolved peak at 1.68 Å associated with 1HS, indicating an ordered distribution of water molecules around SO42-. Similar RDFs were generated for Cl- and Br- systems, with peaks located at 2.22 and 2.40 Å, respectively (Fig. S12b, c). On the contrary, the first peak of the ClO4--H pair is much weaker and less distinguished (Fig. S12d), indicative of some disorganization within the ClO4- solvation shell. The calculated PMF values (derived from RDFs) reflect the kinetic barrier of water molecules in the 1HS, and follow the sequence ClO4-<Br-<Cl-<SO42-. The relatively weak interaction between ClO4- and water molecules leads to facilitated migration and a more disordered distribution of water molecules. Figure S13 illustrates the distribution of the angle θ between two vectors, one from the water oxygen atom to the anion and the other bisecting the water molecule. The probability of θ in the SO42- electrolyte shows two sharp peaks at around 68° and 120°, while water around ClO4- has a broader angle distribution, demonstrating an obvious difference in the degree of disorder, in agreement with the RDF results.

Beside the important function of anions discussed above, the cation (Zn2+) owning double charge should be also responsible for the water structure inside the electrolyte system. Therefore, the interaction between Zn2+ and H2O molecules was studied. Interaction energy of Zn2+ with H2O inside different systems demonstrate an order of Zn(ClO4)2 > ZnBr2 > ZnCl2 > ZnSO4 (Fig. S21a), indicating a larger destroying ability to the original water structure. The RDF and corresponding coordination number from MD simulations displaying the distribution situation of the O atom inside H2O molecule around the Zn2+ were then collected, as shown in Fig. S21b–d. Similar results can be found that Zn2+ inside the Zn(ClO4)2 system possesses the strongest intensity and the largest coordination number with H2O compared with other electrolyte systems. The above results can be explained by the weakest interaction between Zn2+ and ClO4-, as exhibited in Fig. S22.

Solid-liquid transition behavior and tetrahedral entropy analysis

Before an in-depth discussion of the different responses in terms of water structure with respect to ion-specificity and the corresponding antifreeze ability, the Tf of different electrolytes should first be determined. We applied an in-situ observation system based on optical microscopy coupled with a digital camera and a cooling stage to monitor the freezing process of different electrolytes (Fig. 3a and S14). As the temperature was lowered from room temperature at a rate of 1 °C min−1, water drops (50 μL) containing SO42-, Cl-, Br- and ClO4- with a Zn2+ concentration of 5 m (rationale for the choice given in Fig. S15) began to freeze at −23.6, −45.5, −49.5 and −83.5 °C, respectively. The Tf decrease in the anion sequence (ClO4-<Br-<Cl-<SO42-) matches the previous results. Differential scanning calorimetry (DSC) can determine the thermodynamic change that accompanies temperature (Fig. S16) and the temperature sequence for melting peaks is consistent with the observed Tf above. Figures S17 and S18 show representative snapshots of the ice crystal growing process by MD simulations for pure water and the four electrolytes at different temperatures (−20 and −80 °C). At −20 °C (close to the measured freezing point of −23.5 °C), it should be noted that following sufficient equilibrium relaxation and subsequent 150 ns simulation, the ice crystal growth is obvious and ion diffusion is sluggish in the SO42--based electrolyte (Movie S1). However, the ClO4--based electrolyte at the same temperature remains in the liquid state and the ions diffuse uniformly through the entire simulation box (Movie S2). The behavior of Cl- and Br- under the same conditions is similar to ClO4-. As the temperature was further lowered to −80 °C, only the ClO4--based electrolyte system remained in the liquid state, which is consistent with the earlier freezing point results.

Fig. 3: Theoretical and experimental analysis of thermodynamic properties of water in different electrolytes.
figure 3

a In situ optical microscopic observations of four electrolytes (5 m) before and after freezing. Black scale bar, 1 mm. b The probability distributions of the tetrahedral order parameter Qtet for water molecules in different electrolytes at 25 °C. Inset shows the schematic models of liquid-like and ice-like water. c Fraction of ice-like water (tetrahedrality above 0.8) for four electrolytes at 25 °C. d Tetrahedral entropy of water molecules in the four electrolytes, which follows the order SO42- < Cl-<Br-<ClO4- over the temperature range −80 to 40 °C.

Full size image

The discussion of ion-specificity and freezing point at this juncture has largely drawn on qualitative analysis. To gain greater insight into the underlying mechanism, thermodynamic parameters with quantitative values are required. As noted above, one water molecule forms HBs with another four nearby water molecules, some of which possess quasi-tetrahedral structure. In order to resolve the local water tetrahedral degree, the tetrahedral order parameter (Qtet) can be proposed as37:

$${Q}_{tet}=1-\frac{3}{8}\mathop{\sum }\limits_{i}^{3}\mathop{\sum }\limits_{j=i+1}^{4}{\left[\cos {\varPsi }_{jik}+\frac{1}{3}\right]}^{2}$$
(2)

where Ψjik is the angle formed by molecule k and its nearest neighbors i and j. For an ideal tetrahedral structure, this order parameter equals 1 and for ideal gas state, it reduces to 0. Generally, Qtet of water molecules is distributed between 0 and 1 (Fig. 3b). On this basis, we regard the water molecule with Qtet above or below 0.8 as ice-like or liquid-like water, respectively. As indicated in Fig. 3c, the fraction of ice-like water in the four electrolytes follows the order of ClO4-<Br-<Cl-<SO42-, all of which are much lower than that in pure water (Fig. S19). Note that the ice-like water can provide ice nucleation sites, it can be inferred that the ClO4--based electrolyte offers the least possibility for promoting the freezing process. Therefore, the freezing point of electrolytes should be strongly related to the water structure (especially Qtet distribute situation), which is believed to correlate with the entropy of water.

For different aqueous electrolytes, there exist different ratio of tetrahedral structure formed by solely H2O, which should be quantified by a more universal parameter. The total entropy of water can be expressed as:

$$S={S}_{{id}}+{S}_{{ex}}$$
(3)

where Sid is the ideal gas entropy, Sex is the excess entropy and composed of spatial and orientational distributions in liquid systems. For liquid water, the orientational ordering is vital to decide the total entropy of the system due to the significant variation of the structural ordering of water as a result of the nature of the solute. Kumar et al. have employed the variable Qtet to define orientational contribution from the total excess entropy that derives from local ordering of water molecules, known as tetrahedral entropy, which can be written as33

$$\,{S}_{{Q}_{{tet}}}={S}_{0}+\frac{3}{2}{k}_{B}{\int }_{\!\!0}^{1}{{{{{\rm{ln}}}}}}\left(1-{Q}_{{tet}}\right)P\left({Q}_{{tet}}\right)d{Q}_{{tet}}$$
(4)

where S0 is a constant, kB is the Boltzmann constant, P (Qtet) represents the distribution of Qtet at a given temperature T. It can be used to describe the deviation degree of the total S of the system from ideal tetrahedral structure of the water molecule network. With a more negative \({S}_{{Q}_{tet}}\), the severer deviation of the average water distribution from tetrahedral structure.

The tetrahedral entropy \({S}_{{Q}_{tet}}\) was calculated using Eq. (4) based on the population of the tetrahedral order parameter \({Q}_{tet}\) in different systems. Obviously, the pure water exhibits the lowest tetrahedral entropy at room temperature [\(({S}_{{Q}_{tet}}-{S}_{0})/{k}_{B}=-1.770\), Table S1], and the order of \({S}_{{Q}_{tet}}\) in the four electrolytes follows the sequence of ClO4->Br->Cl->SO42-, in accordance with the population of ice-like water shown in Fig. 3c. The sequence remains unchanged at different simulated temperatures ranging from −80 to 40 °C (Fig. 3d). Strikingly, the Tf of water derived from experiments in different electrolytes resulted in very similar \({S}_{{Q}_{tet}}\) (denoted by star symbols in Fig. 3d), which we proposed as a threshold value. The traditional colligative view ignores an ion-specific effect and fails to explain why different electrolytes (at the same concentration) often exhibit distinct frost resistance. The \({S}_{{Q}_{tet}}\)parameter provides a new understanding at a micro-scale that better accounts for and predicts the solid-liquid transition of various electrolytes. Three additional Zn salts with different anions (Zn(TFSI)2, ZnI2 and Zn(NO3)2) were further introduced to verify the reliability of the above proposal. Combining both experimental observation (determining the freezing points as −40.0 °C, −43.0 °C and −62.0 °C for Zn(TFSI)2, Zn(NO3)2 and ZnI2, respectively) and theorical calculation, the \({S}_{{Q}_{tet}}\)for Zn(TFSI)2, ZnI2 and Zn(NO3)2 at each individual Tf generated values close to those obtained for the previous four Zn salts (Figs. S23 and S24). These results validate the use of the threshold value of \({S}_{{Q}_{tet}}\) to predict Tf. Both MD simulations and experimental results confirm that differences in Tf of different salt solutions with unique anions are strongly related to the \({S}_{{Q}_{tet}}\) of water molecules.

The concentration of electrolyte is an important factor related to the Tf, which should be studied further. The most proper concentration of 5 m for Zn(ClO4)2 was sifted according to the DSC results, as presented in Fig. S15. However, the variation of concentration will greatly determine the number of free H2O molecules and also the proportion of ice-like water even at their Tf (Fig. S25). Therefore, it is hard to judge the Tf of any Zn-based electrolyte with random concentration via the \({S}_{{Q}_{tet}}\) value. To verify the versatility of the \({S}_{{Q}_{tet}}\) value for judging the Tf of different Zn-based electrolytes with a same concentration, DSC experiments and MD simulation were further conducted for 1 m and 3 m samples. It can be found that the Tf value of several electrolytes with different anions overall follow well with the order similar to the situation of 5 m, as exhibited in Fig. S26. The \({S}_{{Q}_{tet}}\) under the above Tf were further calculated, as indicated in Fig. S27. Rather close values can both be achieved under the concentration of 1 m and 3 m with different electrolytes, solidly confirming the universality of this judging strategy for Zn-based electrolyte owning different anions but the same concentration.

Low-temperature performance of PANI||Zn batteries

Due to the outstanding antifreeze capability exhibited by the ClO4--based electrolyte, it has been identified as a competitive candidate for practical applications in low-temperature batteries. Apart from preventing electrolyte freezing, high ionic conduction at low-temperature should also be taken into account. In the case of 5 m Zn(ClO4)2 electrolyte, liquid-like water predominates with less HB due to water-water interaction when compared with ZnSO4, ZnCl2 and ZnBr2. In Zn(ClO4)2, the lesser limitation from both the ion solvation shell and bulk electrolyte environment permits greater migration of anions and cations. The ionic conductivities curves of the four standard electrolytes (5 m ZnX, X = SO42-, Cl-, Br-, ClO4-) and one additional electrolyte (ZnTFSI) over the temperature range −80~+25 °C are shown in Figs. 4a and S28. The ZnSO4 electrolyte exhibits a fast decay of ionic conductivity at −20 °C due to freezing. The decay speed varied at lower temperatures for Zn(TFSI)2 with an abrupt decrease at −40 °C. By comparison, Zn(ClO4)2 exhibited the highest ionic conductivity over −80~+25 °C, which agrees with the simulated MSD results and calculated self-diffusion coefficients (Fig. S29). Owing to the lower solid-liquid transition temperature, the 5 m Zn(ClO4)2 electrolyte exhibited sufficient ionic conductivity (1.12 mS cm−1) at the ultralow temperature (−80 °C). This can be ascribed to weak cation interaction with ClO4-, resulting in a dispersed solvation structure, as shown in Fig. S22. The weak coupling of cation and anion affords more free ions to migrate, which guarantees high conductivity. Taking an overview of the reported electrolyte systems with remarkable anti-freezing ability19,38,39,40,41, the 5 m Zn(ClO4)2 used in this work outperforms all others, notably at the ultra-low temperature of −80 °C (Fig. S30). This unique property brings great performance of full cells under an extreme low-temperature environment.

Fig. 4: The electrochemical performance of PANI||Zn full batteries and self-power systems at different temperatures.
figure 4

a Ionic conductivities of the four electrolytes over the temperature range of −80~+25 °C. b The capacity and charge-discharge curves of PANI||Zn full batteries at 1 A g−1 and temperatures in the range 25 °C to −80 °C. c Rate performance at different current densities (0.1 to 1 A g−1) at −70 °C. Cycling performance (1 A g−1) at (d) −70 °C and (e) −80 °C. f Low-temperature performance comparison of this work with previously reported ZIB systems. g Schematic illustrating the self-powered system combining ZIB and TENG for powering a space rover under extremely cold conditions. h TENG-charge and galvanostatic discharge performance of the self-powered system at −80 °C.

Full size image

Since the performance of the full cell is mainly determined by the cathode, it is essential to choose a candidate with sufficient capacity even under ultralow temperature conditions. For traditional inorganic cathodes including Mn, V and Prussian blue (or analogues) based-materials, sluggish Zn2+ kinetics can be attributed to the high associated charge, especially at low temperatures. Consequently, organic materials with charge storage capability involving surface coordination with ions have been studied intensively for use in low temperature batteries19,42,43,44. PANI is a commonly used cathode for ZIBs that can store/release Zn2+ through dynamic coupling/decoupling processes, as illustrated in Fig. S31. In theory, PANI can retain fast ion kinetics even at low temperatures and was accordingly chosen in this study. The capacity and charge-discharge curves of PANI||Zn full batteries utilizing Zn(ClO4)2 as electrolyte at 1 A g−1 and varying temperature are shown in Fig. 4b and S32a. The system exhibited remarkable low-temperature tolerance and high-capacity retention (67.5%) at −40 °C. Three fabricated PANI||Zn pouch cells trapped in ice and operated in series can successfully power LED sets (Fig. S32b). Even at the ultralow temperature of −80 °C, the battery retained a high capacity (74.17 mAh g−1) at 1 A g−1 (Fig. 4b). Fig. 4c demonstrates excellent rate performance ranging from 0.1 A g−1 to 1 A g−1 at −70 °C. Ultra-stable cycling performance over 5000 cycles with ~100% capacity retention and over 1200 cycles with ~85% capacity retention were achieved at −70 °C and −80 °C, respectively (Fig. 4d, e). This level of performance can satisfy the demand for large-scale application under extremely cold conditions, such as polar regions and in space exploration. The performance achieved in this study in terms of working temperature, current density, capacity and cycling stability is compared with existing ZIBs15,19,36,45,46,47,48,49 and other batteries21,50,51,52,53,54,55,56 in Figs. 4f and S33, respectively. Our fabricated ZIBs delivered the best low temperature performance with respect to frost resistance and cycling stability, exhibiting enormous potential in extreme environments. The schematic shown in Fig. 4g illustrates a potential future application scenario for TENG charged antifreezing ZIBs. The TENG can be mounted on top of a space rover and harvest mechanical energy from natural wind in other planets to charge our special ZIBs to supply more stored power under extreme low temperatures. As validation, one ZIB was charged by TENG20 and discharged at a constant current density of 0.4 A g−1 at −80 °C for 2 cycles with satisfactory results (Figs. S34 and 4h). This demonstrates the feasibility of constructing this self-charging power pack with operation in an extremely cold environment.

Discussion

In this work, we quantitatively related an entropic contribution to the Tf of water molecules from a consideration of ion-specific effects. Through in-depth and systematic studies combining multiple experimental methods and theoretical simulations, the HB structure and critical kinetic and thermodynamic data for water molecules in different electrolytes are determined. The water from the ClO4--based electrolyte system with intrinsically higher tetrahedral entropy exhibits the best frost resistance. The Tf of water molecules in different electrolytes can be unified on the basis of a common tetrahedral entropy threshold region. The results demonstrate the crucial role played by tetrahedral entropy and provides a novel approach to understanding water freezing behavior. Guided by this theory, we developed an ultralow temperature aqueous PANI||Zn battery (with a high electrolyte ionic conductivity of 1.12 mS cm−1 at −80 °C), exhibiting a high capacity (74.17 mAh g−1) at −80 °C and 1 A g−1 with a capacity retention of ~85% after 1200 cycles. Moreover, a much better cycling life of 5000 repeats with ~100% capacity retention at −70 °C has also been achieved. Our fabricated battery can be charged using a TENG at an ultralow temperature (−80 °C) and discharged with a stable rate of 0.4 A g−1, confirming feasible application as a self-powered system under extreme conditions. The principles that underpin this work can inform future exploration of the thermodynamic fundamentals of phase transition in designing next generation antifreezing electrolytes for practical application.

Methods

Materials

Zn foils (>99.99%) of 100 μm were used for full batteries construction. PANI was grown on carbon paper by ammonium persulfate oxidizing aniline monomer in aqueous acid57. First, 3.65 mL aniline monomer was added into 150 mL HCl (1 M), then carbon papers were soaked in the solution and keep stirring for 1 h. After that, 150 mL 1 M HCl containing 2.28 g ammonium persulfate was poured into above solution under an ice bath, and the mixed solution was keep stirring for another 1 h. Finally, the carbon papers with dark green PANI were taken out, washed with deionized water and dried at 60 °C for 8 h. The mass loading of the PANI is around 1.1 mg cm−2.

Fabrication of the ZIBs

Commercial Zn sheets (with a thickness of 100 μm and diameter of 10 mm) were utilized as anodes. Before the fabrication, Zn sheets were washed by acetone, ethanol and deionized water in succession. One pierce of Zn sheet anode and PANI cathode were used as two electrodes for constructing the ZIB. 75 μL 5 m Zn(ClO4)2 electrolyte were added into the CR2032 coin cell separated by a piece of glass fiber (GE-Whatman, 125 mm).

Fabrication of the TENG

The rotor is an epoxy glass fiber disk board which has sixteen pieces of 80 μm thick fan-shaped fluorinated ethylene-propylene (FEP) copolymer film stuck on it. The thickness of the rotor is 0.3 mm and the diameter is 20 cm. The stator is a custom-made printed circuit board (PCB) which has 16 pairs of copper (Cu) electrodes. There is a small vertical gap (<0.3 mm) between the rotor and the stator. Two stripes of artificial wool cloth with a width of 2 cm were stuck on a PMMA disk substrate to make triboelectrification brushes.

Simulation methods

Molecular dynamics simulations were performed with GROMACS58 both in the isothermal-isochoric (NVT) and the isothermal isobaric (NPT) ensembles. In both cases, a leap frog integrator with a time step of 1 fs was used. A 1.0 nm cutoff distance were used for the van der Waals interactions and Electrostatic interactions. Electrostatic interactions were calculated using the particle mesh Ewald method59 (PME). In the NVT simulations, V-rescale60 thermostat with a characteristic time of 2 ns implemented the constant temperature condition. In the NPT simulations, 1 atm pressure was maintained via an isotropic Parrinello Rahman barostat. For simulations without ice slab, the dimension of the cubic box (with periodic boundary conditions) is 4.5 × 4.5 × 4.5 nm3, consisting 3025 water molecules, 275 Zn2+ and 275 anions (for SO42-) or 550 anions (for Cl-, Br or ClO4-). The OPC361 water model was adopted, and the interaction parameters for ions were taken from GAFF and Merz62 force field.

The MD simulation for different electrolyte systems from 243 K to 313 K at 1 atm were performed for 60 ns to ensure reaching equilibrium, and the last 10 ns simulation in the equilibrium state were used for analysis. The calculation of hydrogen bond (HB) is based on the geometrical configuration that the distance of two O is <3.5 Å and the angle of O–H···O is <30°. For simulations with initial ice slab, the dimension of the simulation box (with periodic boundary conditions) is 2.8 × 2.3 × 6.9 nm3. The initial ice slab consisted 864 water molecules and the liquid slab contained 600 water molecules, 55 Zn2+ and 55 anions (for SO42-) or 110 anions (for Cl-, Br- or ClO4-). The TIP4P/ICE water model was used in this two-phase system. This water model is especially well suited to perform MD simulations with ice crystal63. The parameters of ions were the same as those above. The two-phase systems were first equilibrated at NVT (using V-rescale thermostat for 1 ns) and NPT ensemble (using Parrinello-Rahman barostat for 1 ns), and then a 150 ns production run at 253 K were conducted.

Potential of mean force (PMF) analysis

The PMF is related to the RDF using the equation PMF(r) = −RTln(g(r)), which reflects the kinetic barrier of water molecules in the first solvation shell.

Mean square displacement (MSD) analysis

MSD measures the mean of the square of particle displacement over time t. MSD(t) = <|r(t)−r(0)|2>.

Self-diffusion coefficient D of water molecules

The self-diffusion coefficient D of water molecules was derived according to the \({{{{{\rm{Einstein}}}}}}\,{{{{{\rm{equation}}}}}}\,D=\frac{1}{6}{{{{{\mathrm{lim}}}}}}_{t\to \infty }\frac{MSD(t)}{t}.\)

Orientational relaxation time τOR of water in first solvation shell of anions

The mean orientational relaxation time of water in first solvation shell of anions are calculated by using the autocorrelation function of orientational order parameter

<P2 (t) P2 (0)>, where \({P}_{2}(t)=\frac{1}{2}(3{{{\cos }}}^{2}\left(\delta \theta \left(t\right)\right)-1).\delta \theta \left(t\right)=\theta \left(t\right)-\theta \left(0\right)\) represents the angle between dipole vectors of a water molecule at times t and 0.

Characterizations

DSC was carried out in METTLER TOLEDD DSC3 in the procedure of +20~−80 °C with a cooling rate of 5 °C min−1, constant temperature for 2 mins and −80~+20 °C with a heating rate of 5 °C min−1. Raman spectroscopy was conducted on Horiba LabRAM HR Evolution microscope and a 532 nm excitation laser was used. FTIR spectroscopy was carried out on FTIR-850 (Tianjin Gangdong Sci.&Tech). The LF-NMR analyzer (NMI20-030H-I, Shanghai Niumag Corporation, Shanghai, China) was used in the measurements of 2D T1-T2 relaxation. The in-situ microscope images of freezing process were obtained by a commercial high-resolution digital camera (Mshot, MS60) equipped with a cooling stage (Linkam THMS600). A direct current motor (XCWS57-30W, WCDQ) was used as the simulated mechanical energy source. The basic performance of TENG was measured by an electrometer (6514, Keithley) and an oscilloscope (MDO3104, Tektronix) with a high-voltage probe (500 MΩ).

Electrochemical measurements

The rate performance and cycling stability of ZIBs were conducted by the Neware battery test system (CT-4008Tn-5V10mA-164, Shenzhen, China). The cutoff voltages were set as 0.5 V and 1.5 V, separately for all electrochemical tests. For rate performance, different current densities of 0.1, 0.2, 0.5 and 1 A g−1 were used while for long-term cycling test, 1 A g−1 were chosen. The discharge–curves of the TENG-ZIBs system were conducted by CHI660E. All low-temperature performances were conducted in HARCICRY refrigerators (Q/22, −15~25 °C) and Jiesheng ultra-low temperature refrigerators (DW-86W28, −86~−40 °C).