Unveil the origin of voltage oscillation for sodium-ion batteries operating at −40 °C

Significance

Sodium-ion batteries (SIBs), as an emerging alternative to Li-ion technology, still face challenges of severe performance degradation at low temperatures. Voltage oscillation at subzero in SIBs has been a common but overlooked scenario, almost yet to be understood. Herein, we demonstrate the origin and elimination strategy of the voltage oscillation and construct a Na-ion full battery with high capacity retention and excellent cycle stability at −40 °C. This work provides combined experimental characterizations and theoretical analyses perspective to re-examine the low-temperature electrochemistry of SIBs and might provide an effective mitigation strategy for other cathodes or electrolyte systems with voltage oscillation at subzero, instead of being limited to Na₃V₂(PO₄)₃ (NVP) or Na₃V₂(PO₄)₂F₃ (NVPF) in SIBs.

Abstract

Voltage oscillation at subzero in sodium-ion batteries (SIBs) has been a common but overlooked scenario, almost yet to be understood. For example, the phenomenon seriously deteriorates the performance of Na₃V₂(PO₄)₃ (NVP) cathode in PC (propylene carbonate)/EC (ethylene carbonate)-based electrolyte at −20 °C. Here, the correlation between voltage oscillation, structural evolution, and electrolytes has been revealed based on theoretical calculations, in-/ex-situ techniques, and cross-experiments. It is found that the local phase transition of the Na₃V₂(PO₄)₃ (NVP) cathode in PC/EC-based electrolyte at −20 °C should be responsible for the oscillatory phenomenon. Furthermore, the low exchange current density originating from the high desolvation energy barrier in NVP-PC/EC system also aggravates the local phase transformation, resulting in severe voltage oscillation. By introducing the diglyme solvent with lower Na-solvent binding energy, the voltage oscillation of the NVP can be eliminated effectively at subzero. As a result, the high capacity retentions of 98.3% at −20 °C and 75.3% at −40 °C are achieved. The finding provides insight into the abnormal SIBs degradation and brings the voltage oscillation behavior of rechargeable batteries into the limelight.

Sodium-ion batteries (SIBs) as attractive energy storage devices have been extensively studied in academia and industry, and the room-/high-temperature performances have been greatly improved (1, 2). However, their operations in extreme conditions (e.g., subsea, military, and defense devices) are limited by insufficient energy/power densities under low temperatures (low-T) (3, 4). The battery deterioration below freezing is mainly attributed to slow Na⁺ diffusion and sluggish charge transfer kinetics. The former includes Na⁺ diffusion in liquid electrolytes and solid electrodes, while the latter involves the desolvation and migration of Na⁺ through a solid electrolyte interphase layer (5). Previous impedance studies show that, instead of ion diffusion, the interfacial process contributes to the bulk of battery impedance at low-T (6). As the temperature decreases, batteries suffer notably enlarged interfacial impedance and severe capacity decay, which is strongly associated with the ion desolvation process (7, 8). Thus, it is well accepted that the key to realizing high-performance SIBs at subzero lies in lowering Na⁺ desolvation energy (9).

Liu et al. improved the desolvation kinetics by using 1, 3 dioxolane/1,2-dimethoxyethane solvents with smaller Li⁺/solvent binding energy (10). Xia et al. designed a weakly solvating solvent to facilitate the desolvation process and thus enable graphite anode to charge in low-T Li-ion battery (LIB) systems (11). We have also constructed weak solvation effects to accelerate the interfacial charge transfer process by reducing electrolyte concentration or introducing liner ether-based solvents (12, 13). However, a conundrum has arisen during the study. In commercial propylene carbonate/ethylene carbonate (PC/EC)-based electrolyte at −20 °C, the voltage oscillation can be observed in the charge/discharge curves of Na₃V₂(PO₄)₂F₃ (NVPF) and Na₃V₂(PO₄)₃ (NVP) cathodes. The degradation of the battery could be alleviated by reducing the salt concentration to 0.3 mol L⁻¹, but the voltage oscillation still exists (12). Such voltage oscillation not only causes unstable output voltage but also leads to severe capacity loss, especially when operating at high current densities or ultralow temperatures, which is also observed in the other PC/EC-based system (14). Unexpectedly, we find that voltage oscillation can be eliminated by replacing the ester-based solvent with ether-based diglyme, enabling better low-T battery performance (13). As known, the temperature dependency of the cell voltage can be quantified by temperature coefficients (TCs), which have a significant association with ion solvation structures (15–17).

Inspired by this, in this work, the correlation between volt-age oscillation and electrolytes has been established to reveal the origins of voltage oscillation in a PC/EC-based system. Based on the theoretical and experimental analysis, the relationship between voltage oscillation and local phase transition is first confirmed by exploring the influence of electrolytes and temperature on the phase transition of the NVP cathode during charge/discharge. It is found that the voltage oscillation in a PC/EC-based system originates from the local phase transition of NVP at −20 °C, which is strongly linked with the exchange current density (j₀) controlled by the interfacial kinetics. The higher desolvation energy barrier in the NVP-PC/EC system would reduce j₀ and then aggravate the local phase transformation, resulting in observable voltage oscillations. Introducing diglyme solvent with lower Na-solvent binding energy at subzero could eliminate the voltage oscillation observed in NVP. As a result, the NVP||Na maintains capacity retention of 99% after 1,000 cycles at 1C at −20 °C, and the NVP||NTO battery exhibits negligible capacity loss after 900 cycles at −20 °C and extra 500 cycles at −40 °C. This work highlights the common but overlooked phenomenon of voltage oscillation and provides different ideas to regulate the electrochemical behavior at low-T.

Results and Discussion

The NVP cathode shows high reversibility at 25 °C in both PC/EC-based and diglyme-based electrolytes (Fig. 1A and SI Appendix, Fig. S1), but lots of “waves” are observed in the voltage curves of NVP in PC/EC-based electrolyte at −20 °C (Fig. 1 A–D). The voltage oscillation of NVP in PC/EC-based electrolyte at −20 °C even can persist in subsequent cycles (Fig. 1E). In contrast, voltage oscillation cannot be found in diglyme-based electrolyte at −20 °C. This voltage oscillation triggers NVP in PC/EC to show much larger polarization (408 vs. 77 mV) between charge/discharge platforms than that in diglyme (Fig. 1F). The polarization will increase sharply with the decrease of temperature (1,190 mV for PC/EC-based electrolyte and 67 mV for diglyme-based electrolyte at −40 °C), leading to the failure of the battery in PC/EC-based electrolyte at lower temperatures (SI Appendix, Fig. S2). As the increase of current density, the voltage polarization in PC/EC-based electrolytes will worsen at higher current densities (>2C) (SI Appendix, Fig. S3). In contrast, the diglyme-based system shows comparable rate capability and long-term cycle stability. The capacity at −20 °C is 98.3% of room-temperature capacity (RTC) and the discharge capacity is about 1.5 times higher than in PC/EC after 1,000 cycles (SI Appendix, Figs. S2–S4). Within 2.5 to 3.8 V, NVP shows a pair of oxidation/reduction peaks at 3.45 and 3.24 V in diglyme at −20 °C (Fig. 1G). However, the anodic peak of NVP at high potential could not be detected completely in PC/EC within 2.5 to 3.8 V, the voltage window of NVP is widened to 2.5 to 4.5 V, in which an oxidation peak at 3.71 V and two reduction peaks at 3.11/2.88 V are detected. Similar voltage oscillation was observed in both half- and full-batteries in PC/EC-based electrolytes at −20 °C (SI Appendix, Fig. S5) (12), indicating that voltage oscillation found at low-T might not be original from Na metal and still exists without Na metal.

Fig. 1.

The charge/discharge curves at 0.1C of the NVP cathode in diglyme- and PC/EC-based electrolytes at (A and B) 25 °C and (C and D) −20 °C. (E) The voltage–time curves and the enlarged views of NVP in PC/EC-based electrolytes at 0.2C at −20 °C. (F) Polarization of NVP||Na battery in both electrolytes obtained at various temperatures. (G) CV curves at 0.1 mV s⁻¹ of NVP in diglyme- and PC/EC-based electrolytes at −20 °C.

What factors contribute to the voltage oscillation of NVP in PC/EC? The ionic conductivity of both electrolytes at −20 °C is practically similar, although the viscosity of PC/EC-based electrolyte is higher than that of diglyme-based electrolyte (SI Appendix, Figs. S6 and S7). Thus, the apparent performance gap between both electrolytes with similar ionic conductivity rules out the possibility that the decreased ionic conductivity is a key factor in voltage oscillations in PC/EC-based electrolytes. To better quantify the temperature dependency of the electrode potential, TCs α are introduced (calculated based on Eqs. 1 and 2). As revealed by the fitting results for the TC of open circuit voltage (OCV) in Fig. 2 B and D (based on the collected isothermal measurements, Fig. 2 A and C), the OCV of Na||NVP batteries shows a downward trend with the decrease of temperature in both electrolytes, indicating positive TCs. Both systems show larger OCV-TCs within 25 to 45 °C (~0.380 mV/K for Na||NVP in PC/EC and 0.325 mV/K in diglyme). Below 25 °C, the OCV is more stable with the changing temperatures, and the OCV-TCs of both systems are 0.074 and 0.134 mV/K, respectively. That is, within −20 to 25 °C, the OCVs of Na||NVP batteries are more stable in PC/EC electrolyte. TCs of OCVs are measured in relative equilibrium, and the entropy contribution is mainly from the electrolytes. The entropy contribution from the NVP phase changes and interfacial charge transfer process should be further considered. Therefore, the TCs based on the median voltage of Na||NVP in PC/EC and diglyme during charge/discharge are further fitted. Their TCs are about 10.744 and 0.128 mV/K, severally in the range of −40 to 25 °C (SI Appendix, Fig. S8). It means that the average voltage of the PC/EC-based system fluctuates greatly with temperatures, which is ~84 times that of the diglyme-based system. Due to the obvious difference between the results of OCV TCs and the average voltage TCs of NVP in both electrolytes, the influence of electrolytes on the voltage oscillation cannot be revealed directly, which needs to further explore the interaction between NVP and both electrolytes, including phase changes, interfacial chemistry, and diffusion kinetics at −20 °C.

Fig. 2.

(A) Isothermal measurement of a Na||NVP in PC/EC coin cell [0% state of charge (SOC)] voltage under varying temperatures and (B) corresponding fitting result for the TCs of OCV. (C) Isothermal measurement of a Na||NVP in diglyme coin cell (0% SOC) voltage under varying temperatures and (D) corresponding fitting result for the TCs of OCV.

According to a previous study, the voltage curve would vary with Na concentration within NVP (18). Thus, the structural evolution of NVP at 25 °C and −20 °C is investigated by in/ex situ X-ray diffraction (XRD). The in situ XRD patterns collected at 25 °C show that the NVP cathode presents a typical first-order phase transformation with a stable intermediate phase (Na₃-xV₂(PO₄)₃) in both electrolytes (Fig. 1A and SI Appendix, Fig. S9 A and B, detailed analysis in SI Appendix), which is consistent well with their similar voltage curves. In contrast, their structural transformation at −20 °C is different, as shown in the ex situ XRD patterns (Fig. 3 and SI Appendix, Figs. S9 and S10). At −20 °C, the intermediate V₂O₃ (labeled as ▼) is detected in both systems, but it appears earlier in PC/EC-based electrolyte, found at C-30% (charged to 30% capacity), while observed at C-60% in the diglyme-based electrolyte (Fig. 3B and SI Appendix, Fig. S10). In addition, some diffraction peaks of the NVP phase [e.g., (104), (113), and (211)], which should be fully transformed into NaV₂(PO₄)₃ after C-95% at 25 °C, disappear at C-60% and reappear at C-100% in PC/EC-based electrolyte (Fig. 3B and SI Appendix, Fig. S10). This phenomenon indicates that NVP in PC/EC presents a discrete phase transition at −20 °C rather than the uniform first-order phase transition. Although the NVP phase can still be detected even after fully charged in both electrolytes along with some diffraction peaks of Na₃-xV₂(PO₄)₃, the discrete phase transition is not obvious in the diglyme-based system.

Fig. 3.

(A) In situ XRD patterns of the NVP cathode in PC/EC-based electrolyte at 25 °C. (B) the color-coded, temperature-resolved, intensity distribution plots correspond to the ex situ XRD patterns and (C) the corresponding charge–discharge curve of the NVP cathode in PC/EC-based electrolyte at −20 °C. (D and E) Schematics showing the phase transitions of NVP at 25 °C and −20 °C upon charging.

Comparing the structural transformation of NVP at 25 °C and −20 °C in PC/EC-based electrolyte, it is found that NVP gradually converts into NaV₂(PO₄)₃ with the extraction of Na⁺ and completely disappears when the extraction amount of Na⁺ arrives at 1.8 at 25 °C (Fig. 3D). Only NaV₂(PO₄)₃ phase could be detected at C-100% state. This NaV₂(PO₄)₃ phase will gradually recover to NVP after the Na⁺ insertion during discharging (Fig. 3 A, Right). In contrast, NVP exhibits in-complete and local phase transitions at −20 °C, and the NVP phase can also be detected even after fully charged, along with some diffraction peaks of Na₃-xV₂(PO₄)₃ (Fig. 3E). Zhou et al. have discussed the origin of voltage oscillations and suggest that the discrete nature of multiparticle phase-separating reactions should be responsible for the oscillatory phenomenon in Li₄Ti₅O₁₂ (19). Thus, the local phase transition found in the NVP-PC/EC system may take major responsibility for the fluctuation of the voltage curve at −20 °C. Although both the Li₄Ti₅O₁₂ anode in LIBs and the NVP cathode in SIBs show similar local phase transition mechanisms and voltage oscillations, the former is the original from the multiparticle properties of the electrode. However, the latter is triggered by low-T, which is a big difference between the two systems.

Since the NVP cathode displays similar electrochemical behavior as well as phase transition at 25 °C in both electrolytes, what triggers their different structural evolution at −20 °C? William C. Chueh et al. point out that the j₀ is closely related to the phase transition process, and the lower j₀ is more likely to trigger local phase transition, leading to voltage oscillation (20). Meanwhile, j₀ reflects the intrinsic kinetics of electron transfer coupled to ion solvation/desolvation and ion transfer (21). Thus, the interfacial chemistry and kinetics of NVP in both electrolytes should be considered. The interfacial chemistries of NVP formed in PC/EC at RT and low-T are carried out as shown in SI Appendix, Figs. S11–S14. The corresponding X-ray photoelectron spectroscopy (XPS) spectra show that the activated NVP in diglyme-based electrolyte exhibits higher content of NaF and Na₂CO₃, while the cycled NVP in PC/EC shows a more serious decomposition of NaPF₆ (detailed in SI Appendix, Figs. S11 and S12) (22, 23). Time of flight secondary ion mass spectrometry (TOF-SIMS) of cycled NVP in both electrolytes at −20 °C (SI Appendix, Fig. S14 A–D) shows that the surface of the cycled NVP is chemical multilayered with organics mainly on the outer layer and inorganics on the inner layer. As shown in SI Appendix, Fig. S14E, the interface layer found on the NVP surface in PC/EC-based electrolyte is porous and jagged with a thickness of 20 to 45 nm, while the CEI layer formed in diglyme is more compact and uniform with a thinner thickness of 20 to 30 nm. Compared with the original solvent-derived organic/inorganic bilayer interphase model of PC/EC-based system, there are more inorganic–organic species in NVP-diglyme system with a relatively high content of Na₂CO₃ and NaF.

Interfacial chemistry is believed to be the key factor limiting low-T battery performance (24–26). Although there are differences in compositions and structure between the CEI formed in PC/EC-based and diglyme-based electrolytes, it is not the main cause of voltage oscillation based on the results of the designed crossover experiment (Fig. 4 and SI Appendix, Fig. S15). Specifically, NVP was activated for 10 cycles in both electrolytes to form a (cathode–electrolyte interphase) CEI layer (marked as PC/EC-CEI and diglyme-CEI), then these batteries were reassembled by replacing the main electrolyte with another one (Fig. 4 D and H). When the bulk electrolyte is PC/EC, no matter whether NVP is activated in diglyme or PC/EC, its subsequent charge–discharge curves are wavy with large polarization (Fig. 4 A, F, and G). Moreover, the diglyme-derived CEI is not conducive to improving the low-T performance of the NVP in PC/EC system. When the bulk electrolyte is diglyme, the voltage oscillation of the NVP cathode with PC/EC-CEI would disappear in several cycles (Fig. 4 B, C, and E). Moreover, the PC/EC-derived CEI layer should make the rate performance of diglyme-based system become worse, especially at high rates, while the diglyme-derived CEI layer has little effect on the rate performance of the PC/EC-based system (SI Appendix, Fig. S15). These results reveal that the voltage oscillation in the NVP-PC/EC system is less associated with the chemical composition of the CEI layer.

Fig. 4.

The first GCD curves of (A) NVP-PC/EC and (B) NVP-PC/EC-diglyme, and (C) the GCD curves of NVP-PC/EC-diglyme after 50 cycles, (D) is the illustration of the crossover experiment converted from PC/EC-CEI layer to diglyme-CEI layer, the first GCD curves of (E) NVP-diglyme and (F) NVP-diglyme-PC/EC, and (G) the GCD curves of NVP-diglyme-PC/EC after 50 cycles, (H) is the illustration of the crossover experiment converted from diglyme-CEI layer to PC/EC-CEI layer.

In addition to the interfacial chemistry, the kinetic mechanisms of NVP in both electrolytes at low-T should also be considered. Typically, the following half-cell reaction happens on NVP, including Na⁺ desolvation and Na⁺ diffusion (solvated Na⁺ + e⁻ + host → Na⁺/host + released solvents/anions). Since the ion desolvation behavior is greatly determined by the electrolyte solvation structure, the solvation structure of both electrolytes and its effect on interfacial kinetics are explored in detail.

Fig. 5 A and B show the snapshots of the molecular dynamics (MD) simulation cell of PC/EC- and diglyme-based electrolytes at 25 °C. Both electrolytes exhibit the solvent-separated ion pair structure at 25 °C (Fig. 5 C and D) (27), in which the first Na⁺ solvation shell is dominated by solvents, and the average coordination numbers of PF₆⁻ are 0.39 and 0.26 per Na⁺, respectively. As the temperature decreases, more anions occupy the first solvation shell (r is ~3.5 Å) in the diglyme-based electrolyte, and the Na-anion coordination number increases from 0.26 at 25 °C to 1.39 at −20 °C (Fig. 5E and SI Appendix, Fig. S16). Therefore, diglyme-based electrolyte exhibits a contact-ion pair structure at subzero (28). In contrast, the ester electrolyte shows an opposite trend. The Na-anion coordination number decreases with the temperature reduction. The relationship between Na-solvent binding energy and coordination numbers as the temperatures change is elucidated by the MD simulation (Fig. 5F). The Na-solvent binding energy increased from −65 KJ mol⁻¹ (25 °C) to −82 KJ mol⁻¹ (−60 °C) in PC/EC-based electrolyte. However, the diglyme-based electrolyte presents no response to temperatures and exhibits similar Na-solvent binding energy at subzero as 25 °C. SI Appendix, Fig. S17 shows the proposed desolvation behavior of both PC/EC- and diglyme-based electrolytes with desolvation barriers of −67.6 and −48.3 kJ mol⁻¹, respectively. The relevant solvation energy ∆G_solv of the two different electrolytes is also measured by constructing a special H-cell (29). By measuring their cell potential (E_cell) at an open circuit, the differences in Na⁺ solvation-free energies in the two electrolytes could be obtained. With 1.0 M NaPF₆ in PC as a reference electrolyte, the measured cell potentials (E_cell) of 1.0 M NaPF₆ in PC/EC and 1.0 M NaPF₆ in diglyme are about -30 and -160 mV, respectively (SI Appendix, Fig. S18A). The calculated solvation energy ΔG_solv of the two electrolytes are about 2.894 and 15.438 kJ mol⁻¹ (SI Appendix, Fig. S18B), inferring that compared with the PC/EC solvents, the Na⁺ shows weaker binding with diglyme solvent, consistent well with the simulation results.

Fig. 5.

The relationship between solvation structure and temperature. Solvation structures of Na⁺ in the electrolytes: snap-shots of the MD simulation cell of (A) PC/EC- and (B) diglyme-based electrolytes at 25 °C; calculated radial distribution functions [g(r), solid lines] of Na⁺-PF₆⁻ in (C) PC/EC- and (D) diglyme-based electrolytes obtained at various temperatures. (E) MD simulation results on the coordination number as the temperatures change. (F) MD simulation results of Na-solvent binding energy and coordination numbers as the temperatures change. (G) Schematic illustrations showing the interfacial process near the electrode/electrolyte interface of the NVP cathode in PC/EC- and diglyme-based electrolytes. Note: E_a for the NVP cathode in both electrolytes is calculated based on the temperature-dependent Nyquist plots according to the Arrhenius equation.

The interfacial kinetics and Na⁺ diffusion mechanism of NVP in both electrolytes are further probed by temperature-dependent electrochemical impedance spectroscopy (EIS) and cyclic voltammetry (CV) methods. The activation energies (E_a) of charge transfer and Na⁺ pass through CEI are both calculated based on the fitting results of the EIS (SI Appendix, Table S1) in diglyme-based electrolyte. The corresponding values of the NVP are ~180 meV and ~80 meV, which are both lower than those in PC/EC-based electrolyte, indicating the reduced Na⁺ desolvation energy barrier and faster interfacial kinetics in diglyme-based electrolytes (Fig. 5G and SI Appendix, Fig. S19C). Moreover, faster Na⁺ diffusion and higher reversibility are revealed by the CV method in SI Appendix, Fig. S19 D and E. Therefore, as revealed by MD simulation and experimental data, the NVP-PC/EC system presents sluggish interfacial kinetics due to the higher desolvation barrier of PC/EC-based electrolytes at low-T.

Based on the above results and the previous reports about voltage oscillation, it could be speculated that the NVP cathode presents voltage oscillation in PC/EC-based electrolyte at −20 °C due to the local phase transition (Fig. 3) (19). Since j₀ is inversely proportional to the discrete phase transition, lower j₀ is more likely to trigger the local phase transition, resulting in voltage oscillation (20). Meanwhile, j₀ reflects the intrinsic kinetics of electron transfer coupled to ion solvation/desolvation and ion transfer (21). Thus, voltage oscillation is closely related to j₀, while j₀ is mainly controlled by the binding energy of Na⁺ and solvents at low-T. Fig. 6 illustrates the relationship among interfacial kinetics (including desolvation and charge transfer), j_0, and voltage oscillation. The higher desolvation energy barrier in NVP-PC/EC system would reduce j₀ and then aggravate the local phase transformation, resulting in observable voltage oscillations. This explains why voltage oscillation in PC/EC-based systems can be mitigated effectively by reducing the desolvation energy, which is achieved by tuning the solvent to diglyme (12).

Fig. 6.

The illustration of the relationship between interfacial kinetics and voltage oscillation.

Based on the above results, the elimination of voltage oscillations lies in regulating j₀, which is highly related to interfacial kinetics. Accelerating interfacial kinetics, including electrolyte solvation structure regulation and interface engineering, could eliminate voltage oscillations at low-T, not limited to NVPF/NVO cathodes or PC/EC-based electrolyte systems. Thus, this strategy could also be applied to other systems (e.g., LiNi_0.8Co_0.15Al_0.05O₂ || Li, NaCrO₂) with voltage oscillations at low-T (30, 31). This also explains why some modified cathodes show no voltage oscillation in PC/EC-based electrolytes (32, 33). The carbon coating or interface modification could effectively accelerate the interfacial kinetics, thus the local phase transition at low-T can be inhibited.

By eliminating voltage oscillations, competitive electrochemical performance could be achieved in the diglyme-based system even at −40 °C. Most of the reported cathodes (33–41), including Prussian blue and its analogs, transition metal compounds (NaTMO₂), and NASICON compounds (e.g., NVP, NVPF), exhibit relatively low capacity retention of RTC at low-T. The discharge capacity retention of the NVP cathode in the diglyme-based electrolyte at −20 °C is about 98.3% of RTC. Even at −40 °C, 75.3% capacity retention of RTC is still retained (Fig. 7A and SI Appendix, Fig. S7). NVP||NTO (sodium titanates) full battery is further fabricated to evaluate the competitiveness of diglyme-based electrolyte. The capacity is calculated based on the mass of NVP. Fig. 7B shows the charge/discharge curves of NVP||NTO at various current rates at −20 °C. A high capacity of 89 mAh g⁻¹ at 0.2C with an average voltage of 3.0 V is achieved. In addition, the battery exhibits a fast-charging ability. When the current density is increased by 50 times, the discharge capacity can still retain 41% at 10C. Achieving long-term cycle stability at low-T is another challenge for SIBs, especially in full configurations. This NVP||Na maintains capacity retention of 99% after 1,000 cycles at 1C at −20 °C, which is superior to the most reported cathodes (33, 35, 40–47) (Fig. 7D). Furthermore, the NVP||NTO battery exhibits negligible capacity loss after 900 cycles at −20 °C. At −40 °C, capacity retention of 99% is still achieved after another 500 cycles (Fig. 7C), showing great superiority in long cycle stability (32, 44, 45, 48) (Fig. 7E).

Fig. 7.

(A) Comparison of capacity retention compared with RT performance for reported cathodes (detailed data are listed in SI Appendix, Table S2). (B) The charge/discharge curves of the NVP||NTO battery in diglyme-based electrolyte at various rates at −20 °C. (C) Cycle stability of NVP||NTO battery in diglyme-based electrolyte at −20 °C and −40 °C. Note: the specific capacity of the NVP||NTO battery is calculated based on the mass of positive active material. (D and E) Comparison of low-T cycle stability compared with other reported (D) half-cell and (E) full-cell (detailed data are listed in SI Appendix, Table S2).

Conclusions

In summary, the intrinsic mechanism of voltage oscillation in the PC/EC-based system at low-T has been revealed. The results indicate that the voltage oscillation found in PC/EC-based system at low-T may be a manifestation of the local phase transformation, which is mainly controlled by Na⁺ desolvation process. The stronger binding energy between PC/EC and Na⁺ at subzero leads to a higher desolvation energy barrier in NVP-PC/EC system. This would reduce j₀ and then aggravate the local phase transformation, resulting in observable voltage oscillations. Therefore, the introduction of diglyme solvent with lower Na-solvent binding energy at subzero could eliminate the voltage oscillation observed in NVP. As expected, NVP in diglyme-based electrolyte delivers superior low-T performance in both half- and full-cells. The high capacity retention is achieved with 98.3% at −20 °C and 75.3% at −40 °C. Moreover, the NVP||NTO battery exhibits negligible capacity loss after 900 cycles at −20 °C, and 99% capacity retention is still achieved after another 500 cycles, even at −40 °C, which is superior to the most reported cathodes.

Materials and Methods

Materials Preparation.

The NVP material is purchased from SHENZHEN KEJING CO., Ltd, China. The sodium titanate (NTO) was prepared via simple calcination treatment from hydrated sodium titanate (Na_0.33H_1.67Ti₂O₅‧1.5H₂O). Na_0.33H_1.67Ti₂O₅‧1.5H₂O was synthesized based on our previous study (49). Typically, the commercial titanium powder (purchased from Aladdin, 99.5% metals basis) was successively ultrasonic in acetone, alcohol, and distilled water for 30 min, respectively. Subsequently, the titanium powder was rinsed repeatedly with distilled water several times and dried for use. Second, 0.3 g of cleaned titanium powder was uniformly dispersed in 80 mL 10 mol L⁻¹ NaOH aqueous solution in a 100 mL Teflon-lined autoclave and heated at 180 °C for 24 h. After cooling down to room temperature, the sample was collected and washed with deionized water and ethanol several times. Then, NTO was obtained by annealing the precursors at 450 °C for 2 h in a mixed H₂ (5%)/Ar atmosphere.

Characterization.

A Bruker XRD detector (scan increment: 0.01°, time per step: 0.76 s) with Cu K_α1 radiation (λ = 1.5418 Å) is utilized to investigate the crystal structure of the NVP powder and electrodes. Scanning electron microscopy (S-4800 15 kV) is used to investigate the morphology of the NVP cathode after cycling. XPS (Thermo Scientific K-Alpha+) is utilized to collect information on the atomic chemical states of the NVP cathode. TOF-SIMS (PHI nanoTOF II) is used to determine the species of atomic units near the surface (about 40 nm depth; sputtering time is ~3 nm/min SiO₂ depth) of the NVP cathode and then forms 3D-mapping images. The ionic conductivity of the electrolytes is measured by a DDSJ-319L (INESA SCIENTIFIC INSTRUMENT CO., LTD, Shanghai) ionic conductivity meter.

Electrochemical Measurement.

When preparing NVP cathodes, the NVP sample, conductive Super P, and sodium carboxymethylcellulose are dispersed in the proper amount of deionized water with a mass ratio of 7:2:1. Then, the obtained slurry is pasted on Al foils. After drying at 70 °C overnight, the foils are cut into small disks with a diameter of 14 mm, where the loading mass of positive active material is around 2 mg cm⁻². The preparation of the sodium titanates (NTO) anode is similar to that of the NVP cathode. The coin batteries are assembled in a pure argon atmosphere. Sodium metal is used as a counter electrode and reference electrode. Whatman glass fiber (GF-D) is chosen as a separator. The PC/EC electrolyte is 1.0 mol L⁻¹ NaPF₆ dissolved in PC: EC =1:1 (v/v) with 5% FEC. The diglyme electrolyte is 1.0 mol L⁻¹ NaPF₆ dissolved in diglyme. The amount of electrolyte added in each coin-type battery is approximately 200 µL. The NVP||NTO batteries are assembled based on coin batteries in which the capacity ratio of positive and negative active material is 1:1.4. Galvanostatic charge/discharge (GCD) tests are conducted on CR2025-type coin batteries controlled by NEWARE battery testing system. Before testing, the batteries are kept quiet at the same temperature for 2 h. EIS is recorded on an electrochemical workstation (CHI660E) in the frequency range of 0.01 ~ 100,000 Hz. CV was measured by the electrochemical workstation (CHI660E, Chenhua, Shanghai).

Simulation Details and Methods.

MD simulations for the solution system of 1M NaPF₆ dissolved in EC-PC-FEC and diglyme are conducted based on Vienna ab initio simulation package (50, 51). The projector-augmented wave method is used to solve the Kohn–Sham equations (52, 53). The exchange–correlation potential is treated within the generalized gradient approximation parameterized by Perdew, Burke, and Ernzerhof (54, 55). A long-range dispersion correction implemented by Grimme (D3) is utilized to describe the vdW interaction (56). The plane-wave cutoff energy is set to 400 eV, and the gamma point is used for Brillouin zone sampling. To accurately illustrate the salt–solvent interactions, all simulations are run for 20 ps. The binding energy between cations and solvents is calculated by the formula of E_{Binding energy} = E_AB-E_A-E_B, in which E_A is the energy of the isolated solute, E_B is the energy of the solvent model and E_AB is the total energy of the electrolyte.

Calculation of TCs:

$$\alpha =\frac{\partial E}{\partial T}=-\frac{1}{\mathit{nF}}\frac{\partial \mathrm{\Delta }G}{\partial T}=\frac{\mathrm{\Delta }S}{\mathit{nF}},$$ [1]

$$\frac{\partial E}{\partial T}=\frac{\partial {\mathrm{\varphi }}_c}{\partial T}-\frac{\partial {\mathrm{\varphi }}_a}{\partial T}.$$ [2]

As shown in Eq. 1, α is proportional to the entropy change of the full-cell reaction, which could also be calculated as Eq. 2, in which ϕ_c and ϕ_a are the equilibrium cathode and anode potentials, respectively.

Calculation of ΔG_solv:

$$\mathrm{\Delta }G=-zFE.$$ [3]

As shown in Eq. 3, z is the number of electrons transferred and F is the Faraday constant, E is the cell potential (E_cell) at the open circuit of the special H-cell constructed according to ref. 29. ΔG_solv is the solvation-free energy of the two electrolytes.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.