More than 50% of global energy ends up as waste heat, including all manner of human activities, natural systems, and all organisms. Among various technologies for waste heat recovery, thermoelectrics have a simple working structure with the advantages of quiet, reliable, and long service life, but they have small open voltages and low energy conversion efficiency because their working substance is electrons or holes1–5. Thermionic capacitors6–8 and thermo-cells9–11 based on ion diffusion and/or reactions have the advantages of large open voltages, but the generated electricity is intermittent and working modes are usually very complicated. Combining the technology of thermo-cells and regular batteries leads to the discovery of thermal charging cells. It has the advantage of a relatively high average output voltage (0-0.5 V) as compared with thermos-cells, but a large temperature difference between anode and cathode is required that is hardly realized in practical batteries12–14. All the above technologies work under a certain temperature gradient, which is far beyond the equilibrium state. Thus, many of the heat energy is dissipated into the environment and the thermal-to-electrical energy conversion efficiency (ηTTE) is low (usually less than 12% of the Carnot efficiency ηC under a temperature difference of 50 ℃). In addition, the measured output voltage in thermo-cells and thermal charging cells contains the contributions from ion diffusion, interface polarization, and the voltage temperature coefficient of battery. The former two terms are one-off, leading to the rapid decay in the output voltage during working. The term of battery’s voltage temperature coefficient can provide a stable and continuous electric output, but the absolute value is very small (usually less than 10% of the output voltage) that has not been accurately measured yet.
Carnot-like cycle is a near equilibrium process. Thus, the ηTTE can be very high because the heat energy dissipated into the environment is extremely small. Thermally regenerative electrochemical cycle (TREC) is a typical Carnot-like (Stirling-like electrical) cycle. During working, the system is charged at high temperature, and then naturally cooled to low temperature and finally discharged at this temperature. If the temperature coefficient of system’s voltage (α) is negative, the absorbed heat at high temperature can be converted to electricity at low temperature and finally released along with the charged electricity. An absolute larger α value leads to a higher ηTTE. Currently, the |α| in TREC is in the range of 0.74–2.27 mV/K and relative Carnot efficiency (ηTTE/ηC) is in the range of 5.6–25%15–25. However, these electrochemical cells usually have small working voltages (< 0.8 V) and/or poor cycling stability (10–160 cycles at 5% capacity attenuation), which greatly impede their wide applications in energy conversion and storage. In contrast, regular batteries have the advantages of large working voltage (1.4–3.8 V), good cycling stability (400–3000 cycles), fast response speed, and portability26–31 as well as extremely high energy conversion efficiency between chemical energy and electricity32–35. Using regular batteries to harvest waste heat under Carnot-like mode can realize a novel and efficient thermally regenerative battery (TRB) technology, but it has not been realized until today because of the low (0.05–1.46 mV/K) 25,36,37 or the absence of battery’s temperature coefficient.
Herein, we successfully developed a thermally regenerative Zn-ion battery to work under Carnot-like mode to extraordinarily harvest waste heat (Fig. 1a). By introducing a Layered Double Hydroxides (LDH) into the anode reaction, extremely high battery’s temperature coefficient of 2.944 mV/K and ηTTE/ηC of 29.24% (Fig. 1b and Table 1), and extraordinary charge-discharge Energy Efficiency (EE) of 104.11% (Fig. 1c) with a large working voltage (1.49 V) and good cycling stability (up to 650 cycles with a capacity attenuation of 1.93%) are realized when the battery is charged at 50 ℃ and then naturally cooled to 5 ℃ for discharge. This study suggests that TRB is one of the most promising technologies for harvesting waste heat (Table 1), which can effectively collect extra electricity in the daytime and then efficiently provide more than 100% of the charged electricity to users such as electric vehicles at night (Fig. 1a).
The structure of NiHCF/Zn battery is shown in Fig. 1a. Nickel hexacyanoferrate (KNiIIFeIII(CN)6, NiHCF) is taken as cathode, zinc is taken as anode, and KCF3SO3 and Zn(CF3SO3)2 are taken as the mixed electrolyte. The chemical reactions of two half cells are shown in Reaction S1 and S2. In order to realize a Carnot-like mode, a Stirling-like electrical cycle is built. It contains four steps (Fig. 2a). For the first step, the battery is heated to a high temperature with the open circuit voltage (OCV) decreasing to V(TH). For the second step, the battery is charged at this high temperature with the heat energy and electricity stored as chemical energy. For the third step, the battery is cooled to a low temperature with the OCV increased to V(TL). During this step, part of the absorbed heat is converted to chemical energy. For the fourth step, the battery is discharged at this low temperature with all the stored chemical energy released.
Table 1 | Summary of various technologies for harvesting waste heat.
|
Thermo-electrics
|
Thermionic capacitors
|
Thermo-cells
|
Thermal charge cells
|
TREC
|
TRB
|
Temperature gradient across the device
|
Yes
|
Yes
|
Yes
|
Yes
|
No
|
No
|
working under equilibrium state
|
No
|
No
|
No
|
No
|
Near
|
Near
|
ηTTE/ηC
|
5-16%
|
0.03-8.46%
|
0.01-12%
|
5.2-7.25%
|
5.6-25%
|
29.24%
|
Output voltage
|
Stable
|
Unstable
|
Unstable
|
Unstable
|
Stable
|
Stable
|
Output voltage for single-pair device
|
0-0.1V
|
0-0.03V
|
0-0.1V
|
0-0.5V
|
0.21-0.81V
|
1.49V
|
Working lifetime
|
> several years
|
15-1000 cycles at 6% capacity attenuation
|
>100 hours
|
50-100 cycles
|
10~160 cycles at 5% capacity attenuation
|
650 cycles at 1.93% capacity attenuation
|
extra heat exchange
|
Yes
|
Yes
|
Yes
|
Yes
|
No
|
No
|
Battery’s temperature coefficient of voltage (αcell) directly determines how much heat energy can be converted to chemical energy during the electrochemical Stirling-like cycle. It contains the contributions from the reactions at anode (α+) and cathode (α−). Thus, the αcell = α+-α−, where the α+ and α− can be given by the Nernst equation (Eqs. (S3 and S4)). The battery’s temperature coefficient of voltage is carefully measured. Both the α+ and α− increase when increasing K+ and Zn2+ concentrations, which can be well described by the Nernst equation (see Fig. 2f,g). The minimum α+ (-0.826 mV/K) and maximum α− (0.692 mV/K) are obtained when the K+ and Zn2+ concentrations are 0.2 mol/L and 1.0 mol/L, respectively. However, the battery is unstable and has larger polarization when K+ concentration is lower than 0.5 mol/L. Therefore, the NiHCF/Zn battery is assembled using the K+ concentration of 0.5 mol/L and Zn2+ concentration of 1.0 mol/L, leading to a total α of -1.221 mV/K. We further test the battery’s performance. A ηTTE/ηC of 14.60% is obtained when the battery is charged at 50 ℃ and then discharged at 5 ℃, resulting in an EE of 95.05% as compared with the value of 92.01% when the battery is charged and discharged at 5 ℃.
Battery’s temperature coefficient is determined by its entropy (ΔS), which can be significantly changed by modifying ion’s types and concentrations (\(\varDelta S\stackrel{\scriptscriptstyle\text{def}}{=}-{k}_{B}\text{ln}\left(\right[x\left]\right)\)) 20, where kB is the Boltzmann constant and x is the effective concentration of ions. Here, by adding 0.05 mol/L of NiSO4 into the electrolyte, an extra chemical reaction occurs at the surface of Zn/Zn2+ electrode
Zn-2e + xNi2++ySO42−+zOH−+wH2O ↔ ZnNix(SO4)y(OH)z·wH2O (1)
ZnNix(SO4)y(OH)z·wH2O has a typical Layered Double Hydroxides (LDH) structure (shown in Fig. 2b). It has two layers. One layer is ZnNix(OH)z, and another is ZnSO4 and water. Strong hydrogen bonds are existed between these layers. Except Zn2+, the LDH material contains Ni2+ and SO42−, which may give additional contributions to battery’s temperature coefficient.
The LDH material on anode’s surface is characterized and confirmed by various techniques. Scanning electron microscope (SEM) measurements show that numerous nanoscale flakes are observed (Fig. 2c and Extended Data Fig. 1a), which are consistent with the character of layered structure in LDH45,46. Energy Dispersive Spectrometer (EDS) mapping revealed the uniform distribution of Zn, Ni, O, and S elements from 100 nm to 10 µm and the molar ratio of Ni: S close to 1: 1 (Extended Data Fig. 1a). The X-Ray Diffractometer (XRD) pattern matches well with that of Zn3.52Ni1.63(SO4)1.33(OH)7.64·4.67H2O (Fig. 2d) 47. Fourier Transform Infrared Spectroscopy (FTIR) and Raman spectra (Fig. 2e) show that the absorption peaks locate at 400–800 cm− 1, 964 cm− 1, and 1635 cm− 1, which are also consistent with the characteristic vibration peaks of Zn/Ni-OH, SO42− and H2O in LDH materials48–50, respectively.
Compared with the pristine battery, the α+ in the modified battery is slightly decreased to about 0.02 mV/K. It still increases with the increase of K+ concentration, consistent with the Nernst equation (Fig. 2f). However, the α− in the modified battery is greatly different from the previous one (see Fig. 2g). It is increased to 2.120 mV/K, 3.18 times of the pristine one, when the Zn2+ concentration is 0.05 mol/L and K+ concentration is 0.5 mol/L. Furthermore, the α− decreases when increasing Zn2+ concentration, which is completely different from the Nernst equation.
The temperature coefficient of the full NiHCF/Zn-LDH battery is measured at 50% state of charge (SOC). The battery has accurate and rapid response to temperature (Extended Data Fig. 5b). The potentials show linear changes to temperature (Extended Data Fig. 3d) and the slope is the temperature coefficient. It is -0.827 mV/K and 2.116 mV/K for the reactions at the cathode and anode, respectively. Ultimately, the NiHCF/Zn-LDH battery has a very high absolute α of 2.944 mV/K, superior to all the reported electrochemical cells near room temperature (Fig. 2h).
We simulate the Stirling-like cycle for NiHCF/Zn-LDH battery with the data shown in Fig. 3a,b. A charging and discharging range of 15–85% SOC is chosen. First, the battery is charged at various high temperatures but discharged at 5 ℃ under a current density of 11.2 mA·g− 1 (0.2 C). Although the charging temperature is different, the difference in discharge voltage is small with the error bar less than 1.85%, indicating that battery’s discharging performance is rarely affected. However, the charging voltages are significantly reduced when increasing charging temperature. Specifically, we test the battery via charging at 50 ℃ and discharging at 5 ℃ under different current densities. When the current density increases, battery’s discharge voltage gradually increases while the charging voltage decreases.
The maximum P and EE values reach 104.13% and 104.11%, respectively, when the battery is charged under 0.2 C and worked when charging at 50 ℃ and discharging at 5 ℃ (Fig. 3c,d). This is the first reported EE value above 100% for various batteries. Compared with the normal working mode, it can provide 13.26% extra electricity when the waste heat is collected and recovered (Fig. 3e). Thus, the calculated ηTTE/ηC is 29.24%, which is superior to all the values for various harvesting waste heat technologies (see Fig. 1b and Table 1). For example, it requires an average thermoelectric figure of merit (ZT)ave of 2 for thermoelectrics to compete this value, which is far larger than current best thermoelectric materials near room temperature. We also assemble a soft pack battery for performance test. It is charged at 48 ℃ in the daytime and at 5 ℃ in the basement at night (Fig. 3f). At the same capacity, the charging voltage at 5 ℃ is 1.726 V but it is decreased to 1.517 V at 48 ℃ with a reduction of 12.11%. This further strongly suggests that harvesting waste heat can greatly improve battery’s performance. The high ηTTE/ηC can be understood from the point view of working principle of thermal engine. Firstly, all the four processes are nearly isothermal and thus there is no extra heat loss during working. Secondly, low working current density leads to the small Joule heat. Both of them are responsible for the observed high energy conversion efficiency.
The LDH material plays an important role on battery’s performance. We conduct FTIR, Raman, in situ XRD and EDS measurements to investigate the variation of LDH material during charging and discharging processes. In situ XRD data (Fig. 4a) shows the diffraction peak belonging to the Zn-LDH phase gradually appears at 32.8° during discharging, suggesting the generation of the LDH material. Extended Data Fig. 6b shows the Ni and S contents on the electrode surface gradually decrease during charging process, indicating that the LDH is gradually decomposed. The characteristic peaks of H2O, SO42−, and M-OH in FTIR (Extended Data Fig. 6c) and Raman (Fig. 4b) are also gradually strengthened when battery’s capacity is decreased, implying that the formation of LDH material. All the data indicate that the LDH material is obviously involved in battery’s working. It is gradually generated during discharging but decomposed during charging processes.
The above data shows that the chemical reactions at anode contain Reaction S2 and Reaction 1. Thus, the temperature coefficient α− should also have the contributions from these two reactions, which can be given by a modified Nernst equation (see the details in supplementary section 3). The relationship between the total temperature coefficient of these two reactions (αmix) and Zn2+ concentration is given by Eq. S16, which shows an opposite trend with the previous Nernst equation (Eq. S4). The Zn2+ concentration not only influences the ionic entropy, but also affects the occurring proportion of Reaction S2 and Reaction 1. When Zn2+ concentration is high, the αmix is dominated by Reaction S2; however, when Zn2+ concentration is low, the αmix is dominated by Reaction 1 (Extended Data Fig. 6d). The experiment data can be well fitted by the modified Nernst equation (Eq. S16) (see Fig. 2g). High αmix is expected when Zn2+concentration is small and/or the temperature coefficient of Reaction 1 is high. However, very small Zn2+ concentration (< 0.05 mol/L) leads to large polarization and thus low P and EE. Therefore, the Zn2+ concentration is chosen as 0.05 mol/L in this work to realize high performance NiHCF/Zn-LDH battery.
The previous NiHCF/Zn battery has low cycling stability. After the first 20 cycles, the capacity is quickly decreased to 13.89%. However, the modified NiHCF/Zn-LDH battery has much better cycling stability because the LDH material can isolate water from Zinc electrodes. The galvanostatic charge/discharge (GCD) test at 1C shows that the battery has an attenuation rate of 1.93% after 650 cycles (Fig. 4c). In addition, the reversible capacity in rate performance test can recover to 96.63% when the current switches back to 0.1 C from 3C (Extended Data Fig. 7h). Table 1 summarize the performance of various technologies for harvesting waste heat. TRB shows great advances as compared with the others, and thus is one of the most promising waste heat recovery technologies.
In conclusion, we have successfully developed thermally regenerative Zn-ion battery to efficiently collect and recover the waste heat. The very high energy conversion efficiency and excellent battery’s performance indicate that harvesting waste heat by TRB is very powerful and useful. This strategy is expected to be extended to other batteries such as Li-ion, Na-ion, and K-ion batteries in the future.