Introduction

Cooling is essential to modern human civilization1,2, and its importance is expected to escalate owing to global warming. Paradoxically, current active cooling technologies, mainly driven by the vapor compression cycle, consume huge amounts of energy and contribute significantly to global warming3,4. In contrast, passive cooling, which functions without active energy input, could address this cooling challenge5,6,7,8. Passive cooling technologies have shown great promise in cooling buildings9,10, solar farms6,11, and electronics12,13, as well as sub-ambient pharmaceutical storage14, among others. To this end, water—the most mysterious substance on Earth—could play an indispensable role. Water evaporative cooling, one of the most promising passive cooling methods, leverages the enthalpy of vaporization absorbed during water evaporation to generate cooling15. The evaporative cooling power density is a product of the enthalpy of vaporization and the evaporation flux of water16, which, in turn, is a product of the mass transfer coefficient and the vapor pressure difference (between ambient and evaporation surface)17. Typically, evaporative cooling performance is primarily tuned by changing the evaporation rate.

Recent studies have claimed the tunability of the enthalpy of vaporization using hydrogels, which are water-rich polymeric materials. Various hydrogels based on poly(vinyl alcohol) (PVA) have suggested reductions in the enthalpy of vaporization of water18,19,20. The lower enthalpy was attributed to the presence of a layer of intermediate water between the bound water on the polymer chains and the free bulk water21. Intermediate water is a layer of water bound by weak hydrogen bonds. Hence, it was argued that vaporization enthalpy is lower when there is a larger ratio of intermediate water to free water. Raman spectroscopy is a useful method for qualitatively assessing the content of intermediate water by monitoring the relative intensities of the peaks around 3600 cm−1 within the OH bond stretching frequency band21,22. Hydrogels are found to exhibit higher relative intensities at this range than liquid water. Hydrogels with lower evaporation enthalpies also have higher relative intensities than other hydrogels22. It has been proposed that the reduction of enthalpy resulted in superior steam generation performance when using specialized hydrogels. Besides steam generation, hydrogels are promising materials for passive evaporative cooling applications6,17. Due to the interactions between hydrophilic polymer chains and water molecules, large amounts of water (more than 90 wt%) can be retained in a jelly-like solid structure (also known as ‘solid water’)21,23. This enables evaporative cooling of surfaces without the risk of water spillage. Not only are hydrogels easier to handle than liquid water, but they can also be molded into various shapes to cool irregular, inclined, or vertical surfaces such as food containers and building walls10,14.

Besides controlling the behavior of water molecules through new materials, water molecules also respond strongly to electrostatic fields thanks to their strong polarity. The most common demonstration of this characteristic is the bending of a tap water stream toward a rubber balloon (or any electrical insulator) that has been electrostatically charged by friction24. Molecular dynamic studies have shown enhancement in water vaporization rates when electrostatic fields are applied perpendicular to the water surface25,26. As water molecules are polarized and aligned vertically, the weakened intermolecular attraction enables more molecules to break away from the hydrogen bond network26. Our group has also demonstrated experimentally that a vertical electrostatic field enhances solar steam generation by 15.6%27. Supplementary Raman spectroscopy at three different water-dominant modes (translation, OH bend, and OH stretch) also suggests weakened hydrogen bond networks under electrostatic fields. As a result, it has been deduced that the electrostatic field reduced the enthalpy of vaporization for water27. Nevertheless, whether there are changes in evaporative cooling power (the product of enthalpy and evaporation rate) remains an open question, considering the simultaneous increase in the evaporation rate and the supposed drop in the enthalpy of vaporization.

Another mechanism for electric field-enhanced water evaporation is through ionic or corona winds, i.e., electrohydrodynamic (EHD) enhancement. When a sharp needle is supplied with high voltages, partial air breakdown occurs, and the gas molecules in its vicinity are ionized28. If a ground electrode of significantly larger radius of curvature (flatter) is placed nearby, the ions are attracted to the ground electrode by Coulomb force. The ions collide with neutral air molecules as they travel across the electrode gap, transferring momentum to the air molecules29. The result is an ionic wind flowing from the sharp, high-voltage corona electrode to the flatter ground electrode. Ionic winds have been exploited to enhance heat convection in electronics cooling30,31,32,33, food drying34,35, and even as aerospace propulsion36. Most electronics cooling applications have focused on maximizing ionic wind output at high voltages, enhancing convective heat transfer on the surfaces of high-power electronics. In the field of food drying, ionic winds have the advantage of achieving a high drying rate with considerably lower energy requirements than conventional convective (fan-driven) and oven drying37. Besides energy savings, the texture and nutritional contents of food are better preserved since the food products are not heated to high temperatures as in oven drying37. Studies of ionic wind food drying largely center around the drying rate of food products. In contrast, the influence of ionic wind on evaporative cooling power is not well investigated38. Additionally, in these conventional applications, cooling performances around the corona onset voltage range are poorly understood31,32.

To benchmark cooling performances, the Coefficient of Performance (COP) is typically adopted to evaluate the efficacy of cooling systems. It is the ratio of the total cooling power generated to the total energy consumed39. Since there is some energy expenditure, albeit minimal, in the electrostatic field-enhanced evaporative cooling, this should lead to a finite COP. Therefore, it is appropriate to quantitatively gauge electrostatic field-enhanced cooling via its COP.

Herein, we find that electrostatic field can serve as a tool to tune evaporative cooling power, which results in significant enhancement of water-based evaporative cooling, as illustrated in Fig. 1. In the threshold region (around 5 kV) of corona wind generation, evaporative cooling power can be enhanced by 88% with only a 3% input of electric power. Through systematic investigation, ionic wind is proven to be a major driving force contributing to the enhancement. While increasing voltage leads to a continuous rise in evaporation rates, the enhancement in evaporative cooling power is most pronounced in the voltage range around the threshold voltage of corona (i.e., the onset of corona discharge). As ionic current and wind velocities increase at high voltages, increased heat convection from the environment offsets the net cooling effects. The electric power in the circuit also surges at high voltages, diminishing the energy efficiency of the process. In the threshold region, the COP of this enhancement mechanism far exceeds that of conventional evaporative coolers, even reaching around 3000 at the lowest voltage due to minimal energy expenditure. To widen potential application of this finding, we extend the study to solid water in hydrogel. Similar enhancements in evaporative cooling performances under electrostatic fields were observed, confirming the potential of electrostatic fields as a means of enhancing passive evaporative cooling in practical applications. Furthermore, it is essential to note that vaporization enthalpy of solid water in hydrogel varies considerably under electrostatic fields, while liquid water shows limited enthalpy change. Raman spectroscopy attributes this to the increase in intermediate water under electrostatic fields. Compared with liquid water, the COP of solid water evaporative cooling is only marginally lower, but it boasts better adaptability and practicality. In summary, our work presents a mechanism for enhancing evaporative cooling in both liquid and solid water, broadening the potential application of water evaporative cooling for a sustainable energy future.

Fig. 1: Schematic illustration of water evaporation and evaporative cooling under electrostatic field enhancement.
figure 1

Localized ionization of air molecules near the high-voltage electrode creates an ionic wind blowing across the water surface, enhancing the evaporation rate. Furthermore, the polarization and alignment of water molecules at the surface via the electrostatic field reduce the energy barrier for evaporation. Resulting from the confluence of these factors is enhanced evaporative cooling.

Full size image

Results

Water evaporation and temperature under electrostatic fields

To obtain empirical evidence of EHD enhancement of evaporative cooling, a simple evaporative cooling setup was constructed, as shown in Fig. 2a. Due to thermal insulation at the bottom and minimal thermal radiation indoors, heat convection can be assumed to be the only significant heat inflow into the system, with evaporative cooling as the only heat outflow from the system. Thus, the evaporative cooling flux (power density), \({q}_{{{{\rm{cool}}}}}\) is given by

$${q}_{{{{\rm{cool}}}}}={m}_{{{{\rm{evap}}}}}{H}_{{{{\rm{vap}}}}}={h}_{{{{\rm{conv}}}}}\Delta T={q}_{{{{\rm{conv}}}}}$$
(1)

where \({m}_{{{{\rm{evap}}}}}\) is the evaporation flux, \({H}_{{{{\rm{vap}}}}}\) is the vaporization enthalpy of water, \({h}_{{{{\rm{conv}}}}}\) is the heat convection coefficient, \(\Delta T\) is the temperature difference between the ambient air and water, and \({q}_{{{{\rm{conv}}}}}\) is the convection heat flux. While a temperature gradient exists between the water surface and the aluminum base, the temperature of the aluminum base can be assumed to be close to that of the water surface due to the small thickness of the water layer. Furthermore, the temperature gradient is minimal during thermal equilibrium, where water temperature is constant and heat fluxes throughout the water body are minimal. Therefore, the temperature readings on the aluminum plate are used to represent the water temperature, although they are at the upper bound of the temperature range within the water, as the lowest temperature typically occurs at the evaporation surface. Thus, the evaporative cooling flux measured is underestimated to some extent. Calculation of \({q}_{{{{\rm{cool}}}}}\) using \({m}_{{{{\rm{evap}}}}}{H}_{{{{\rm{vap}}}}}\) is straightforward under normal circumstances, where \({H}_{{{{\rm{vap}}}}}\) is assumed to be constant. Since \({H}_{{{{\rm{vap}}}}}\) could vary under electrostatic field, \({h}_{{{{\rm{conv}}}}}\Delta T\) was used instead to calculate \({q}_{{{{\rm{cool}}}}}\).

Fig. 2: Measurement of temperature and power input under electrostatic field-enhanced water evaporation.
figure 2

a Schematic of the setup to evaluate the electrostatic field-enhanced evaporative cooling on liquid water. Scatter plots in gray represent 2-min average values, while the black horizontal line segments represent the averages for the entire segments (error bars represent ±standard deviation). b Time-series evaporation flux (\({m}_{{{{\rm{evap}}}}}\)) under stepwise increasing voltages from 0 to 9 kV. c Time-series temperature readings under stepwise increasing voltages and stable relative humidity (RH) of air. d Average evaporation flux and leak current plotted against voltage [Inset: Equivalent circuit showing leak current flow]. e Equilibrium temperature difference between water and ambient air (\(\triangle T\)) and electrical power/Joule heat (\({P}_{{{{\rm{elec}}}}}\)) plotted against voltage. f Enhancement ratio (relative to quantity at 0 V) of \({m}_{{{{\rm{evap}}}}}\) and \(\triangle T\), juxtaposed with \({P}_{{{{\rm{elec}}}}}\) [Inset: Divergence between enhancements of \({m}_{{{{\rm{evap}}}}}\) and \(\triangle T\) under negligible Joule heat]. Data shown in (d–f) are mean ± standard deviation (≥3 measurements).

Full size image

In Fig. 2b, the average evaporation mass flux, \({m}_{{{{\rm{evap}}}}}\), shows a clear increasing trend as the voltage, V, is increased stepwise. The evaporation mass flux plot was derived from the mass plot shown in Supplementary Fig. 1. At the maximum voltage of 9 kV, \({m}_{{{{\rm{evap}}}}}\) is almost 6 times that at 0 V. Figure 2c depicts the water layer temperatures at different applied voltages, where the ambient temperature and relative humidity (RH) are shown for reference. When water was first poured into the setup, evaporation started naturally, and water temperature started dropping due to evaporative cooling, \({q}_{{{{\rm{cool}}}}}\) (see Supplementary Fig. 2). The increasing temperature difference between water and ambient air, \(\Delta T\), causes an increase in convective heat input, \({q}_{{{{\rm{conv}}}}}\). Thermal equilibrium is reached when \({q}_{{{{\rm{conv}}}}}\) matches \({q}_{{{{\rm{cool}}}}}\), and \(\Delta T\) remains constant at 3 °C, as seen at the start of Fig. 2c. As the electrostatic field enhanced \({m}_{{{{\rm{evap}}}}}\), the water temperature dropped further below the initial equilibrium, indicating enhanced evaporative cooling. At 6 kV, \(\Delta T\) increased by 60% to about 4.8 °C. Subsequent voltage increases resulted in slight saturation in cooling performance. As the voltage supply was turned off, \({m}_{{{{\rm{evap}}}}}\) returned to its initial rate, while \(\Delta T\) decreased to its initial equilibrium value. This is a clear indication of the causal link between V and \({m}_{{{{\rm{evap}}}}}\), as well as its associated \({q}_{{{{\rm{cool}}}}}\).

Figure 2d shows that \({m}_{{{{\rm{evap}}}}}\) increases continuously with increasing voltage, which might be attributed to the reduction of \({H}_{{{{\rm{vap}}}}}\) by an electrostatic field, as reported before27. There is also a positive correlation between \({m}_{{{{\rm{evap}}}}}\) and electric current, suggesting possible energy input in the form of Joule heat. Equilibrium \(\Delta T\) plotted in Fig. 2e experiences a significant increase from 3 kV to 6 kV, after which it levels off and starts decreasing slightly. It is worth noting that the decrease in \(\Delta T\) coincides with the increase of leak current from 6 kV onward, hinting again at possible Joule heat involvement. Assuming the copper wiring connecting the electrodes to the electrostatic generator has negligible electrical resistance, total Joule heat is equivalent to the total electric power (\({P}_{{{{\rm{elec}}}}}=I\times V\)). However, as electrical resistance in air is much larger than that in water, the bulk of Joule heating occurs within the air gap between wire and water, as depicted in the inset of Fig. 2d and detailed in Supplementary Note 3. Nevertheless, the power of Joule heating in water is far too low to offset and then reduce evaporative cooling power to level off \(\Delta T\) (see Supplementary Fig. 5a), all while evaporation rates continue rising with increasing voltages. Moreover, if \({H}_{{vap}}\) was calculated using these assumptions, the reduction in \({H}_{{{{\rm{vap}}}}}\) is unreasonably high (~71%) (see Supplementary Fig. 5b). Besides, Fig. 2f provides that \({m}_{{{{\rm{evap}}}}}\) enhancement starts to diverge from the \(\Delta T\) enhancement even with negligible electrical power consumption (and Joule heating) around 3–5 kV. In essence, evaporative cooling performance has been underestimated when it is calculated using \({h}_{{{{\rm{conv}}}}}\Delta T\) by assuming constant \({h}_{{{{\rm{conv}}}}}\). The changes in \({h}_{{{{\rm{conv}}}}}\) under increasing voltages must be ascertained to obtain a more accurate estimation of the cooling power.

Measurement of ionic wind and cooling flux

When electrostatic field was applied, ripples on the water surface could be clearly observed (Supplementary Movie 1), indicating additional kinetic energy input from the field. One possible mechanism of kinetic energy input is via ionic wind. When high voltage is supplied to a thin wire, the air molecules adjacent to the wire are ionized. As the ions are driven toward the ground electrode (i.e., aluminum plate) by Coulomb forces, the ions collide with and impart momentum to neutral air molecules along the way, resulting in a wind blowing from the high-voltage wire to the water surface. Thus, a separate experiment was performed to confirm the involvement of ionic wind in evaporation enhancement. The same parameters were measured when the pathway of ionic wind was blocked using a physical barrier. For more effective wind blocking, the high-voltage wire was replaced with a sharp syringe needle fixed at a higher height (60 mm above the water surface). In Fig. 3a, b, \({m}_{{{{\rm{evap}}}}}\), \(\Delta T\), and current increased when the needle was supplied with high voltage. Upon thermal equilibrium, a hard polystyrene plate was inserted 18 mm directly below the needle tip to completely block the wind path from needle to water surface, while the high voltage was continuously supplied. Polystyrene has low relative permittivity and exerts minimal distortion to electrostatic field strength and distribution, so the field intensity at the water surface can be assumed to be largely unchanged. The blockage led to all three parameters (\(\Delta T\), \({m}_{{{{\rm{evap}}}}}\), and electric current) returning to approximately their initial equilibrium values, as if no voltage was applied. The blockage of the ionic pathway prevented wind from blowing toward the water surface, resulting in no enhancement in evaporation and cooling. Moreover, the recorded electric current decreased to almost zero, confirming that ionic wind is responsible for the observed electric current instead of leak circuit. After the removal of the plate, the prior enhancement and electric current reappeared. Thus, ionic wind was proven to be the main driving force of the evaporation and cooling enhancement under electrostatic fields, and the electrical current detected was indeed ionic currents. For additional proof, a rippling dent on the water surface under the needle was clearly observed when high voltage was applied. When the ionic path was blocked, the ripples disappeared instantly, returning to its initial stillness (details to be found in Supplementary Movie 2).

Fig. 3: Measurement of ionic wind and cooling flux.
figure 3

a Effects of ionic wind blockage on evaporation flux (\({m}_{{{{\rm{evap}}}}}\)) and ionic current. Scatter plots in light green represent 2-min average values for \({m}_{{{{\rm{evap}}}}}\), while the green horizontal line segments represent the averages for the entire segments (error bars represent ±standard deviation) [Inset: Schematic of the wind-blocking setup]. b Effect of ionic wind blockage on water temperature. c Correlation between average measured wind velocity (\(u\)) and ionic current, plotted against voltage [Inset: Schematic of the wind measurement setup]. d Variations of evaporative cooling flux (\({q}_{{{{\rm{cool}}}}}\)) and heat convection coefficient (\({h}_{{{{\rm{conv}}}}}\)) with wind velocity. Dashed lines represent simulated results from COMSOL. Data shown in (c, d) are mean ± standard deviation (≥3 measurements).

Full size image

Wind velocity, \(u\), must be determined to better assess the effects of ionic wind on evaporation and evaporative cooling. The previous cooling experiment with copper wire was repeated in conjunction with wind measurements. Due to the non-uniform velocity and direction of ionic winds, the wind measurements only provide an estimation of the average wind velocity across the entire water surface. Nonetheless, the measured wind velocities showed a strong correlation with electric current in Fig. 3c. Higher ion flow rates caused a higher rate of collisions with air particles, resulting in stronger winds. At low voltages, small electric currents caused the wind velocities to increase only slightly. Corona onset was reached as voltage increases beyond 5 kV, causing both electric current and wind velocity to increase drastically. However, considerable evaporation and cooling enhancements were present even before corona onset.

\(u\) can therefore be used to estimate \({h}_{{{{\rm{conv}}}}}\) and ultimately \({q}_{{{{\rm{cool}}}}}\). At 0 V, the initial cooling power can be calculated as \({{q}_{{{{\rm{cool}}}}}}_{0}={{m}_{{{{\rm{evap}}}}}}_{0}{H}_{{{{\rm{vap}}}}}\), where \({H}_{{{{\rm{vap}}}}}\) of water has the well-known value of 2453.6 J g−1 at room temperature. Subsequently, the initial heat convection coefficient (at 0 V) can be obtained via \({{h}_{{{{\rm{conv}}}}}}_{0}=\frac{\,{{q}_{{{{\rm{cool}}}}}}_{0}}{{\Delta T}_{0}}\), where \({\Delta T}_{0}\) is the temperature difference at 0 V. With Nusselt number relations, \({h}_{{{{\rm{conv}}}}}\) can then be scaled using wind velocity via a correlation of \({h}_{{{{\rm{conv}}}}}\,\propto \sqrt{u}\), as shown in Fig. 3d (\({{h}_{{{{\rm{conv}}}}}}_{{{{\rm{n}}}}}={{h}_{{{{\rm{conv}}}}}}_{0}\sqrt{\frac{{u}_{{{{\rm{n}}}}}}{{u}_{0}}}\))40,41. Subsequently, \({q}_{{{{\rm{cool}}}}}\) across different u can be calculated via \({{q}_{{{{\rm{cool}}}}}}_{{{{\rm{n}}}}}={{h}_{{{{\rm{conv}}}}}}_{{{{\rm{n}}}}}{\Delta T}_{{{{\rm{n}}}}}\). At 0.33 m s−1 (7 kV), the \({q}_{{{{\rm{cool}}}}}\) was 4 times that at 0.06 m s−1 (0 V). It is worth noting that at 7 kV, a cooling power enhancement of about 144 W m−2 was achieved with just 38 W m−2 of electrical power. The enhancement ratio is highest at lower voltages, especially before corona onset. Strikingly, at 3 kV, every 1 W of electrical power could generate almost 500 W of extra cooling power, as illustrated in Supplementary Fig. 15. COMSOL simulations (details in Supplementary Note 6) were also performed to ascertain the increase in \({h}_{{{{\rm{conv}}}}}\) and \({q}_{{{{\rm{cool}}}}}\) values, by increasing the impinging jet velocity in accordance with the formula:

$${u}_{{{{\rm{ion}}}}}=\sqrt{\frac{{Id}}{\rho {bA}}}$$
(2)

where \(I\) is electric current, d is the distance between electrodes, \(\rho\) is air density, b is ionic mobility, and \(A\) is the surface area of the flat plate42. The simulated \({h}_{{{{\rm{conv}}}}}\) and \({q}_{{{{\rm{cool}}}}}\) values show good agreement with the experimental values, corroborating the validity of the scaling correlation.

Application on solid water medium

Next, we extend a similar investigation to solid water in a widely studied PVA hydrogel (details of synthesis and thermal characterization provided in “Methods” section and Supplementary Note 8). This simple form of hydrogel is easy to fabricate, safe and non-toxic, making it ideal for wide-scale usage. To test its cooling capacity under varying electrostatic fields, a similar setup was constructed, as illustrated in Fig. 4a. Comparing solid water in hydrogel with liquid water at 0 V, solid water has slightly higher evaporation flux, as shown in Fig. 4b. This is due to lower \({H}_{{{{\rm{vap}}}}}\) in solid water of 1824 J g−1 measured via differential scanning calorimetry, as depicted in Supplementary Fig. 17. Besides the lower \({H}_{{{{\rm{vap}}}}}\), the surface morphology could also have an impact on the evaporation performance. The porous structure of hydrogels can result in a higher effective evaporation surface area than a liquid water surface, giving rise to higher evaporation43. Nonetheless, the extent to which surface morphology influences evaporation flux remains unquantified. There are also reports suggesting the causal relationship between the effective surface area and evaporation flux is not linear, contrary to conventional understanding44. The lower \({H}_{{{{\rm{vap}}}}}\) also causes the cooling flux in solid water to be lower than that in liquid water. As voltage increases, evaporation enhancement in solid water is more significant than that in liquid water, whilst enhancement of cooling flux is weaker in solid water than that in liquid water. This phenomenon signifies that there is a sharper reduction of \({H}_{{{{\rm{vap}}}}}\) in solid water under electrostatic field relative to that of liquid water, which will be further discussed in Fig. 5.

Fig. 4: Evaporative cooling with solid water in PVA hydrogel.
figure 4

a Schematic of the setup to evaluate the electrostatic field-enhanced evaporative cooling on solid water. b Comparison of cooling (\({q}_{{{{\rm{cool}}}}}\)) and evaporation (\({m}_{{{{\rm{evap}}}}}\)) fluxes between solid and liquid water. Solid markers represent solid water, while hollow markers represent liquid water. c Comparison of the equilibrium temperature difference between water and ambient air (\(\triangle T\)), and electrical power (\({P}_{{{{\rm{elec}}}}}\)) under different voltages. Solid markers represent solid water, while hollow markers represent liquid water. d Time series of IR images (top view) showing differences in temperature distribution between liquid water and solid water in the hydrogel. The vertical line at the center of the images is the wire electrode. e Coefficient of performance (COP) of the system across different voltages, plotted in logscale. The average range of COP for active evaporative coolers is indicated [Inset: Close-up of COP beyond 4 kV under a linear scale]. Data shown in (b, c, e) are mean ± standard deviation (≥3 measurements).

Full size image
Fig. 5: Variations in vaporization enthalpy of water due to changes in molecular bonding networks.
figure 5

a Percentage decline in calculated enthalpy of vaporization (\({H}_{{{{\rm{vap}}}}}\)) of solid and liquid water across different voltages. Data shown are mean ± standard deviation (≥3 measurements). b Schematic of the setup for Raman spectroscopy on solid water (hydrogel). c Raman spectrum of fully hydrated PVA hydrogel, deconvoluted into 1 CH peak and 3 OH Stretch sub-peaks, namely 2 free-water peaks (FW1 and FW2) and 1 intermediate water peak (IW). Relative intensities and peak shifts of FW1, FW2, and IW sub-peaks under increasing electrostatic fields are illustrated separately (Intensities are relative and not inter-comparable among different sub-peaks). d Ratio of peak intensities of intermediate (weakly bonded) water (IW) to free bulk water (FW1 + FW2) under different voltages. The intensities of sub-peaks are represented by two sub-peak parameters, namely peak height (green plots) and peak area (dark red). Solid markers represent solid water, while hollow markers represent liquid water. e Shifts of sub-peak center frequencies (FW1, FW2, and IW) under increasing voltages. Solid markers represent solid water, while hollow markers represent liquid water. f Magnitude and direction of dipole moment of water molecules under varying perpendicular electrostatic fields, obtained via molecular dynamics simulation [Inset: Angle of dipole moment relative to electrostatic field direction]. Data shown are mean ± standard deviation (≥3 measurements).

Full size image

Despite the lower cooling power density, solid water in hydrogel provides substantially higher \(\Delta T\) than liquid water (shown in Fig. 4c), maintaining the bottom surface at a much lower temperature. This is attributed to weaker heat transfer from the ambient air to the aluminum plate, as there is no internal convection within the solid water in hydrogel (which is present in liquid water). The cooling progression of both liquid water and solid water shown in the IR images of Fig. 4d presents a visual representation of the difference in internal convection across the samples. Details can be found in the “Methods” section and Supplementary Movie 3. Throughout the cooling process, the temperature distribution across liquid water is visibly more uniform than hydrogel, as liquid water of different temperatures can readily mix and equilibrate. Hydrogel, on the other hand, shows clear temperature gradient from the center to the sides, as the surface closer to the wire electrode undergo greater cooling. Besides, the internal convection in liquid water causes it to reach thermal equilibrium earlier (at around the 10th minute), while the hydrogel surface continues to cool beyond 10 min. The time taken for liquid water to reach equilibrium (~10 min) also correlates well with the duration shown in the temperature plot of Fig. 2c. In summary, the lower equivalent \({h}_{{{{\rm{conv}}}}}\) explains the comparatively lower cooling flux in solid water despite its higher \(\Delta T\). The superior \(\Delta T\) and thermal insulation of solid water in hydrogel make it ideal for sub-ambient storage of heat-sensitive food and pharmaceutical products14. This is especially impactful in impoverished regions with limited access to refrigeration. The impact of electrostatic field on the evaporative cooling of food items using hydrogel is demonstrated in Supplementary Note 11 and Supplementary Movie 4.

Figure 4e provides the COP of both cooling experiments under different voltages, which is the ratio of cooling power to input power. Active evaporative coolers (using fans and dehumidifiers, etc.) typically have an average COP range of around 10-8045,46. Significantly, the COP in this work at voltages below the corona onset voltage (<5 kV) far exceeds that of active evaporative coolers, providing additional cooling energy with miniscule electrical power. Furthermore, \(\Delta T\) values reach saturation beyond corona onset (>5 kV), as stronger ionic winds increase heat input from the environment via convection. Therefore, pre-corona-onset voltages are more favorable for sub-ambient cooling applications due to high energy efficiency. The combination of a wide temperature control range and miniscule energy expenditure allows for ultra-efficient thermal regulation.

Molecular network rearrangement and vaporization enthalpy

Given the calculated \({q}_{{{{\rm{cool}}}}}\), values of \({H}_{{{{\rm{vap}}}}}\) can be derived via \({H}_{{{{\rm{vap}}}}}=\frac{{q}_{{{{\rm{cool}}}}}}{{m}_{{{{\rm{evap}}}}}}\) as shown in Fig. 5a. Enthalpy reduction in water under electrostatic fields is only marginal, as the initial value is generally within reach of the confidence intervals of subsequent values, which are propagated from the errors in experimental data. Solid water, on the other hand, experiences more significant enthalpy reduction under electrostatic fields, as evaporation enhancement outpaces cooling enhancement. This difference in behavior could be attributed to changes in molecular alignment and bonding networks.

Hydrogen bonds are responsible for the unique properties of water, as they are the primary intermolecular bonds among water molecules. These intermolecular bonds directly influence the intramolecular bonds within water molecules47. For further insights on the bond energy of water networks under electrostatic fields, in-situ Raman spectroscopy was conducted on both liquid water and solid water in hydrogel under different voltages. Figure 5b provides a schematic illustration of the setup for solid water testing. To prevent water loss due to possible Joule heating from the Raman laser, the hydrogel was placed in a plastic container of water, with only its top surface exposed and not covered with water. To facilitate continuous replenishment of water, the bottom surface of the gel was perforated to increase surface area for water uptake. The wire electrode was held at the same height as the previous experiments using spacers. To test liquid water, a drop of water was simply dropped onto the silicon wafer. Detailed setup information and test results are available in the “Methods” section. Under Raman spectroscopy, water molecules generally have 3 main types of vibrational modes, namely translation, OH bending, and OH stretching. These modes correlate to the intramolecular bond energies within water molecules, which can be influenced by the intermolecular hydrogen bond network. In particular, the OH stretching frequency band (3000–3800 cm−1) is especially sensitive to changes in the intermolecular bonding and networks48,49. Since hydrogen bonds are electrostatic attractions between a more electropositive H atom and the more electronegative O atom of the neighboring molecule, a single water molecule can interact with multiple neighboring molecules through various configurations48. The OH stretching frequency band can therefore be deconvoluted into multiple sub-bands, which represent different intermolecular bonding configurations. The OH band obtained from the spectroscopy was deconvoluted into 3 sub-peaks, around 3259 cm−1, 3456 cm−1, and 3605 cm−1. The first 2 sub-peaks correspond to water networks having stronger intermolecular bonds, whereas the 3605 cm−1 sub-peak is associated with molecules with weak or no hydrogen bonds18,48. Adhering to common hydrogel nomenclature, the first 2 peaks are named free-water (FW) peaks, while the 3605 cm−1 sub-peak is named intermediate water (IW) peak. For the solid water spectrum shown in Fig. 5c, the additional 2916 cm−1 sub-peak represents the CH stretching mode of the PVA polymer chains21. Figure 5c also offers a visual depiction of the changes in the relative intensities and peak frequencies of all three water sub-peaks for solid water under increasing electrostatic fields. Detailed analyses are shown in Fig. 5d, e, while details on sub-peak deconvolution are provided in Supplementary Tables 2 and 3.

Most notably, the relative intensity of the IW peak in solid water strengthens with increasing voltage. Figure 5d shows that the IW:FW ratios of peak height and peak area both increase for solid water, while no clear trend can be observed for liquid water. The increase in IW relative intensity can also be clearly seen in Fig. 5c. As the IW peak corresponds to weakly bound water molecules, an increasing proportion of IW would enable molecules to detach and escape more easily from the evaporation surface. This indicates decreasing \({H}_{{{{\rm{vap}}}}}\) with rising voltage in the case of solid water, a finding which is in line with the enthalpy trend in Fig. 5a.

Figure 5e shows a general redshift for the FW2 and IW sub-peaks of solid water, as is also visible in Fig. 5c. While redshifts generally suggest lengthening and weakening of bonds, the peaks in question correspond to the intramolecular OH bond and not the intermolecular H bonds. Instead, this indicates a strengthening of H bonds, as they are negatively correlated with intramolecular OH bond strength. Since water molecules are highly polarizable, the molecular dipole moment induced by the electrostatic field forces the molecular network to align along its direction. To illustrate this effect, a simple molecular dynamics simulation was performed using the rigid TIP4P/2005 water model (details in “Methods” section). Figure 5f shows the effect of electrostatic field on the magnitude and direction of the dipole moment of water molecules. Without electrostatic polarization, water molecules at the interface are prone to aligning parallel to the interface (90° from field direction)50. As electrostatic field is applied and increased, the dipole moment increases in magnitude and aligns more closely with the field direction, as shown by the decreasing angle. The increase in dipole moment enhances the intermolecular bonds in the network, and thus weakens the intramolecular bonds, which explains the redshifts seen in Fig. 5e51. Although stronger intermolecular bonds are typically associated with higher vaporization enthalpies, the alignment effect at the water-air interface are also known to influence evaporation as it greatly impacts surface tension. On the other hand, \({H}_{{vap}}\) is widely accepted to have a positive correlation with the surface tension at the interface52,53. Without polarization, the parallel alignment of molecules at the interface causes the high surface tension of water, effectively forming a barrier preventing molecules from escaping50,54. This orderly interfacial barrier is disrupted when a perpendicular electrostatic field is applied. The dipole moment becomes increasingly aligned with the field, and the molecular network is stretched perpendicular to the interface. This reduces surface tension and enables water molecules to escape more easily55. Therefore, the reduction in \({H}_{{{{\rm{vap}}}}}\) can partly be attributed to the molecular alignments at the surface under electrostatic field. The results from in-situ Raman investigations and MD study on the molecular level suggest that electrostatic field interacts strongly with the H bond network in both liquid water and solid water, with stronger influence on the latter.

In essence, water evaporation is mainly driven by the vapor pressure gradient between the water surface and the ambient air. Ionic wind greatly increases the vapor pressure gradient without changing vaporization enthalpy, yielding enhancements of evaporation rate. In contrast, electrostatic field significantly impacts on the hydrogen network in both liquid and solid water, which tends to lower the vaporization enthalpy, facilitating evaporation. While the slightly lowered vaporization enthalpy curtails cooling capacity, the greatly enhanced evaporation rate boosts cooling capacity. Consequently, cooling power, the product of evaporation rate and vaporization enthalpy, is significantly enhanced. Thus, one can see that electrostatic field acts as a catalyst to increase the cooling power in both liquid and solid water, by increasing rate of convective mass transfer at the air-water interface (via EHD effect) and tuning the vaporization enthalpy (via electrical polarization effect). Akin to chemical catalysts, which greatly enhance reactions without extra energy input, the additional energy incurred by the generation of ionic wind is negligible in comparison to the cooling enhancement achieved, leading to unprecedented COP.

Discussion

In a boost to the adoption of low-carbon passive cooling, electrostatic field-induced water evaporation has been proven to be an effective technology to tune and enhance evaporative cooling. This enhancement arises from the combined effects of ionic wind (the EHD effect) and the tuning of the vaporization enthalpy of water (the electrical polarization effect). The cooling power is amplified as the ionic wind increases air flow near the evaporation surface, thereby boosting the evaporation rate. While higher voltages lead to stronger ionic winds and increased evaporation rates, the optimal voltage range for efficient sub-ambient cooling lies below the corona onset voltage. Within this range, the coefficient of performance (COP) of the system exceeds 1000, significantly surpassing the average COP of industrial active evaporative coolers, which is typically below 100. This enhancement technique has also been demonstrated with solid water in a PVA hydrogel. Due to the lack of internal convection, evaporative cooling with solid water in a hydrogel can achieve even colder sub-ambient temperatures at the cooling surface, despite a lower cooling flux. This characteristic makes hydrogels particularly promising for practical applications in sub-ambient food and pharmaceutical storage. Conversely, the electrostatic field appears to reduce the enthalpy of vaporization of solid water in hydrogel by restructuring the water networks and aligning the molecules at the surface, a finding corroborated by Raman spectroscopy. In contrast, the enthalpy reduction in liquid water under an electrostatic field is less pronounced. While the enthalpy drop partially offsets the cooling output, a significant net gain in cooling power is still generated by the electrostatic field with minimal energy expenditure. At the optimal range, the cooling enhancement far exceeds—by 2–3 orders of magnitude—the electrical power consumed, functioning similarly to a catalyst that accelerates a chemical reaction without requiring additional energy input. This research lays the groundwork for energy-efficient tuning of evaporative cooling in both liquid and solid forms of water, reinforcing the potential of water-based evaporative cooling as a sustainable alternative to conventional cooling methods.

Methods

Setup information

For experiments involving liquid water, 22 mL of water was held on a thin aluminum plate using a thin hollowed-out plastic frame, forming a 3-mm-thick water layer with an area of 85 × 85 mm. A self-adhesive thermocouple (OMEGA Engineering SA1-K-72-SRTC) was attached to the bottom of the aluminum plate to measure its temperature. The bottom surface of the aluminum plate was then attached to a high-density polystyrene foam to minimize heat exchange with the environment. To apply an electrostatic field perpendicular to the evaporating surface, a 0.1 mm-diameter copper wire was suspended at a fixed height of 9 mm above the water surface using an acrylic support frame. The copper wire was connected to the voltage supply terminal of an electrostatic generator (Trek 610E). An aluminum plate was connected to the ground terminal via an electrical multimeter (Keysight 34465 A) for current measurement. To obtain the evaporation rate of water, the setup was placed on a precision weighing balance (Ohaus Pioneer PX523). The sliding doors of the weighing balance were closed to create an enclosure to minimize wind disruptions from the environment.

For experiments involving solid water (hydrogel), a similar setup was used, except the base aluminum plate and plastic frame (to hold water) was instead replaced by a 55 × 55 mm aluminum plate which was small enough for the 78 × 78 mm hydrogel to fully cover. Similarly, the aluminum plate was attached with the same thermocouple and attached to the polystyrene foam. The same HV wire-holding frame was used, with its height adjusted to ensure the high-voltage wire was suspended 9 mm above the hydrogel surface.

Experimental details

To measure wind velocity, a hot-wire anemometer (RS PRO RS-72H) was used. The anemometer sensor was placed 20 mm adjacent to the wire and 12 mm above the water surface to avoid blocking the flow path of ions. The effects of electrostatic field or ion transport on the accuracy of anemometer readings were proven to be minimal (see Supplementary Note 5). For quantities plotted in Figs. 3c, d and 4b, c, each experiment (data point at each voltage) was taken at thermal equilibrium (after water/hydrogel temperature has stabilized) for around 10 min. All quantities were taken at an interval of 1 s. Furthermore, the experiment for each data point was repeated at least 3 times. The average value of each quantity is plotted, while the uncertainty (standard deviation) is derived considering variations both between repetitions and within each repetition (variation within ~10 min). After each experiment (data point repetition), the amount of water lost was replenished to ensure the next experiment would start with the same amount of water. This is done to minimize the variation caused by differences in water level, as is present in the time-series experiment in Fig. 2.

To ensure the accuracy of measurement, the hydration status of the hydrogel is kept similar in all experiments by immersing it in water after several experiments. Nonetheless, variations in water content are inevitable among earlier and later experiments. To understand the impact of this variation in water content, the sequence of the experiments was deliberately arranged to repeat the same data point (voltage) under different water content. For example, if the 5 kV data point was done at the end of a cycle (low water content), it was repeated early in the next cycle (high water content). The variations (errors) in \(\Delta T\) and \({m}_{{{{\rm{evap}}}}}\) resulting from the variations in water content were shown to be insignificant (Fig. 4b, c). In other words, the variations in water content were small enough to avoid noticeable changes in the evaporation and cooling behaviors of the hydrogel.

The error bars represent the standard deviations of experimental values. The standard deviations of derived values further down the line, such as \({q}_{{{{\rm{cool}}}}}\) and \({h}_{{{{\rm{conv}}}}}\), are duly derived via the error propagation method from the standard deviation of upstream experimental data (e.g., \({m}_{{{{\rm{evap}}}}}\), \(\Delta T\), …).

Hydrogel synthesis

Poly(vinyl alcohol) (PVA) powder (Mw 89,000–98,000), with a hydrolysis assay of 99.0–99.8%, was purchased from Sigma-Aldrich. The PVA powder was dissolved in deionized water at a mass ratio of 1:9 (10 g PVA in 90 mL water) and stirred at 95 °C until a homogenous transparent solution was obtained. After the solution was allowed to cool to room temperature, it was poured into a PVDF mold. The solution was then subjected to 3 cycles of freezing at −30 °C (>12 h) and thawing. The result is a solid physically crosslinked hydrogel with high water content. Supplementary Fig. 16 shows the 78 × 78 × 4 mm piece weighing 31.25 g used in the experiments.

Infrared thermal characterization

Infrared (IR) imaging was employed to identify differences in the cooling behaviors of both liquid and solid water media, as shown in Fig. 4d. The FLIR E60 IR camera was used for this purpose. The video recordings (Supplementary Movie 3) were captured from the top as shown in Supplementary Fig. 24. At the beginning of the video, the samples were allowed to reach thermal equilibrium before electrostatic field was applied.

Raman characterization

A WITec alpha300 R Raman spectrometer was used to conduct in-situ Raman spectroscopy on both PVA hydrogel and liquid water. The short magnification lens of \(\times\) 20 was used. The excitation laser used had a wavelength of 532 nm and had a low power of less than 1 mW to minimize Joule heating and water loss.

The spectrum data for water is deconvoluted into 3 different peaks, namely 2 separate free-water peaks (stronger intermolecular bonds) and 1 intermediate water (weaker intermolecular bonds) peak. The spectrum for PVA hydrogel has one more peak around 2915 cm−1, which is attributed to the CH bond of the PVA polymer. Subsequently, the peak area and height ratios of (intermediate water: free water) are computed to ascertain the effects of electrostatic field on the bonding network of water molecules, as depicted in Supplementary Tables 2 and 3.

To identify the range of electrostatic field across the evaporating surface, a COMSOL simulation was performed using the Electrostatic module. Supplementary Fig. 25 shows the model setup and the simulated electrostatic field distribution across the setup when high voltages (2 and 6 kV) are supplied to the copper wire. The correct material properties (including relative permittivity) were assigned to each geometry. While the electrostatic field above the water/hydrogel surface ranges from 4485 V m−1 to 225,105 V m−1, most of the surface is subjected to an electric field of more than 104 V m−1.

Molecular dynamics simulation

A simulation box of \(40\times 40\times 200\) Å was constructed with periodic boundary conditions applied on all directions, in which 1605 water molecules were inserted at the center to simulate a 30 Å thick water slab with a density of \(\rho=1\) g cm−3, as shown in Supplementary Fig. 26. Their initial positions were generated randomly inside LAMMPS while initial velocities were taken from a Gaussian distribution corresponding to T = 50 K. The interaction of water molecules was modeled using TIP4P/2005 model with a cutoff distance rc = 10 Å. The TIP4P/2005 water molecule parameters are shown in Supplementary Fig. 27. The particle-particle, particle mesh (PPPM) method with an accuracy of \(2\times {10}^{-5}\) was utilized to solve for long-range electrostatic interactions beyond the cutoff distance. The intramolecular bonds and angles were fixed using the command “fix rigid” in the large-scale atomic/molecular massively parallel simulator (LAMMPS) package (version: stable_2Aug2023) used to run the simulations. An integration timestep of 0.5 fs was used, and each output used for analysis was printed every 1000 timesteps (0.5 ps) and averaged throughout the whole sampling run.

For each value of electric field in the simulations, the temperature of the system was initially ramped from T = 50 K to T = 300 K in an NVT ensemble for 0.2 ns. After reaching the desired temperature, the system was then equilibrated for another 0.5 ns under the same conditions. Finally, 2.5 ns NVE sampling runs were performed, during which the outputs were sampled for analysis.