Here, we propose fast-switching transparent viologen-based EESDs with dual cathodes exhibiting four distinct EC modes and achieving visible-NIR dual-band EC energy storage functionality. Transparent polyviologen (PV) and WO3 ∙ 2H2O are articulately chosen as cathodic materials on both terminals, ingeniously adopting PEDOT:PSS as a protective and proton-providing layer for polyviologen34. Impressively, PEDOT:PSS not only effectively prevents the polyviologen from being degraded but also acts as a proton source to increase the speed of the reversible redox reaction and enhance the interfacial contact conductivity. Combined with the uniform electric field produced by the UV-cured solid-state electrolyte-coated zinc mesh35, which is highly transparent, the resulting EESDs exhibit high transparency, fast switching, exceptional capacitive contribution, uniform coloration, and strong potential for large-scale manufacturing. The current approach leads to significant energy-saving effects in two ways. First, the device has a high energy storage capacity, allowing it to return the energy used to drive it. Furthermore, the device features self-coloring and self-bleaching capability, achieving zero energy consumption. Second, the dual-band, four-mode EESDs, when used as a smart window for buildings, can regulate the transmission of visible and NIR light, enabling intelligent light-heat management and significantly reducing HVAC energy consumption. Experimental and simulation tests for all-day, year-round, and across the United States demonstrate that smart building windows can effectively regulate the light and heat entering indoor spaces, accounting for 10% energy savings compared to conventional low-emissivity (low-E) glass.

We propose conceptual EESD smart windows with four distinct modes: transparent, visible colored, NIR colored, and fully colored, designed to adapt to different times of the day, seasons, and regional weather conditions. The structure of the polyviologen|Zn mesh|WO ∙ 2HO device is illustrated in Fig. 1. It features a nearly symmetrical configuration, with polyviologen/PEDOT:PSS and WO ∙ 2HO cathodes attached to ITO-PET substrates on the top and bottom, respectively, enabling independent control over visible and NIR light transmission. The central layer consists of a UV-cured solid-state aqueous electrolyte-coated zinc mesh, which serves as the anode. This architecture allows for controllable self-charging (bleaching) and discharging (coloration) independently in visible and NIR regions. When sunlight enters a building, the visible light provides illumination while generating significant heat, whereas NIR light contributes solely to heat gain without enhancing visibility. Thus, our smart window was designed to intelligently regulate these two spectral bands, optimizing indoor brightness and temperature control. In cold weather (Fig. 1a), the WO ∙ 2HO layer remains neutral and transparent, allowing NIR light to enter and heat the interior. Meanwhile, the polyviologen layer can either remain transparent to maximize indoor brightness and warmth or switch to colored mode for privacy protection, functioning as a dynamic "blind". During hot conditions (Fig. 1b), the WO ∙ 2HO layer darkens, blocking NIR light to reduce heat gain. Similarly, the polyviologen layer can be adjusted based on indoor brightness and temperature needs. For example, at midday, when sunlight is intense, polyviologen switches to colored mode to reduce both visible light and heat, maintaining a comfortable indoor environment. On cloudy days, at dawn, or during dusk, it remains transparent to maximize natural lighting. By dynamically modulating sunlight transmission, this smart window effectively reduces the reliance on artificial lighting and HVAC systems, significantly lowering building energy consumption.

In realizing the dual-cathode, four-mode smart window, we envisaged the EESD based on a polyviologen|Zn mesh structure. To achieve this, we specifically designed and synthesized a highly transparent polyviologen, with characterization as detailed in Methods. The sprayed polyviologen thin film on ITO-PET exhibited over 90% optical transmittance, surpassing conventional EC material films (Fig. S1, Supplementary Information). To address the inherent drawbacks of low response speed due to low conductivity and poor stability due to swelling in polyviologen, the conductive polymer, PEDOT:PSS, was introduced as a protective and enhancement layer. The device architecture is illustrated in Fig. 2a, while Fig. 2b displays the morphology and thickness of the corresponding layers, with the polyviologen layer measuring 3.9 μm and the PEDOT:PSS layer measuring 80 nm. The Zn mesh with a wire diameter of 36 μm maintained high transmittance, with an average optical transmittance of 88.5%, as shown in Fig. S2 (Supplementary Information). The electrolyte-coated Zn mesh also retained high transparency, with minor absorption dips at 1437 nm, 1930 nm, and 2500 nm, indicating water absorption (Fig. S3, Supplementary Information). PEDOT:PSS was chosen due to its high electrical conductivity, stability, and low redox potential, mainly benefiting from its relatively low interfacial resistance with aqueous zinc-ion electrolytes. During the charging and discharging process, reversible Zn deposition and dissolution occur at the Zn mesh anode, while the PEDOT and polyviologen cathodes undergo redox reactions, as described by the following electrochemical equations:

The chemical structure changes of polyviologen during this redox process are depicted in Fig. 2c.

Cyclic voltammetry (CV) tests were performed to analyze their electrochemical performance (Fig. 2d). At a scan rate of 2 mV s⁻, the device without the PEDOT:PSS layer exhibits only one pair of a reduction peak at -0.66 V and an oxidation peak at -0.41 V, corresponding to the reversible process transitioning from PV to PV⁺ and back to PV. For the polyviologen|Zn mesh EESD with the PEDOT:PSS layer, besides the oxidation (-0.40 V) and reduction peaks (-0.58 V) of polyviologen, there is an additional minor oxidation (-0.88 V) and reduction (-1.05 V) pair associated with PEDOT:PSS (Fig. 2e). Thus, during reduction, the voltage decreases from the neutral charging state at -1.2 V to -1.05 V, causing PEDOT⁺ to convert into PEDOT. Attributed to the enhanced ionic conduction and the presence of a potential step provided by the PEDOT:PSS layer, faster reduction of polyviologen occurs at -0.58 V, which is reflected in a significant increase in peak current.

At various scan rates from 1 to 10 mV s, both the anode (I) and cathode (I) peak currents increased with the square root of the scan rate but did not show a linear correlation, indicating that the device's behavior is not solely diffusion-controlled (Fig. 2f). The devices with the PEDOT:PSS layer exhibited higher peak currents, with a significant increase in peak current as the square root of the scan rate increased, suggesting an enhancement in the electron transfer rate. Further analysis of the charge storage mechanism was conducted. Based on the equation  , a plot of log(i) versus log(v) was generated, where i and v represent the peak current and scan rate, respectively (Fig. 2g). The calculated b values are approximately 0.85, falling between 0.5 (diffusion-controlled) and 1.0 (surface capacitive control). This value indicates that the PEDOT:PSS layer does not affect the dual control mechanism of charge storage, which is governed by both diffusion and surface capacitance. To further elucidate the contributions of diffusion and capacitive control, the capacitive contribution at various scan rates was calculated. As shown in Fig. 2h, the results reveal that capacitive contributions dominate, increasing from 65.5% to 85.7% as the scan rate increases, with the PEDOT:PSS layer having no significant impact on this behavior. It should be noted that, with such a high capacitive contribution, the current EESDs demonstrate exceptional surface-controlled charge storage and exceptional high-rate performance.

The electrochemical impedance of the device was measured (Fig. 2i), and an equivalent circuit was employed for quantitative analysis of the impedance. In the equivalent circuit, the intersection on the real axis represents the internal resistance (R), while the diameter of the semicircle corresponds to the interfacial charge transfer resistance (R). The results indicate that the presence or absence of the PEDOT:PSS layer does not affect the internal resistance of the electrolyte, which remains around 30 Ω. However, the PEDOT:PSS layer significantly influences the interfacial resistance, reducing it from 1356.7 Ω to 532.3 Ω. This result further confirms the role of the PEDOT:PSS layer in enhancing the electron transfer rate.

The PEDOT:PSS layer enhances the electrochemical performance, leading to a noticeable improvement in the electrochromic properties. When the transmittance was measured at various voltages ranging from -1.5 V to -0.1 V, the PEDOT:PSS layer did not impact the transmittance in its transparent state, which remains around 67.63% (Fig. 3a, b). However, a considerable difference is observed in the colored state, with the device containing the PEDOT:PSS layer showing a transmittance contrast of 64.23%, compared to just 44.41% for the device without it. Photographs of polyviologen|Zn mesh EESDs with PEDOT:PSS during the discharge (coloring) and charging (bleaching) processes are presented (Fig. 3c). The response speed of the two devices also exhibits a notable difference (Fig. 3d). The coloring and bleaching times of the device without the PEDOT:PSS layer are 30.1 s and 38.2 s, respectively. In comparison, aided by the enhanced ion migration and electron transfer of the PEDOT:PSS layer, the device achieved an astonishingly fast switching times of 1.6 s for coloration and 0.8 s for bleaching. To the best of our knowledge, this is one of the few reports in which a viologen-based EC device has achieved a response time of less than one second. The darker color in the colored state results in a higher optical density, thereby increasing the coloration efficiency from 137.10 to 389.44 C cm (Fig. 3e). The PEDOT:PSS layer also offers protective benefits for the polyviologen layer by preventing direct contact with water and zinc ions in the electrolyte, effectively reducing water-induced swelling and ion erosion of polyviologen. This protective effect is evident in the long-term cycling stability test (Fig. 3f). After 6600 s of cyclic operations, the device with the PEDOT:PSS layer maintains 90.58% of its ΔT, while the device without it retains only 57.03%.

Moreover, the PEDOT: PSS layer dramatically enhances the energy storage capacity of viologen-based EESDs by promoting the involvement of more polyviologen molecules in redox reactions. In the discharge test (Fig. 3g), the devices with and without the PEDOT: PSS layer show capacities of 141.9 and 73.2 mAh m, respectively, at 0.5 mA cm. The capacity measurements at varying current densities revealed similarly significant improvements (Fig. 3h). A comprehensive performance comparison between our device and other zinc-based EESDs, presented in a radar chart (Fig. 3i), highlights the competitive overall performances of the polyviologen|Zn mesh EESD with PEDOT:PSS, particularly excelling in response speed and coloration efficiency (Table S1, Supplementary Information).

In contrast to the zinc foil, the zinc mesh offers the benefit of preserving transparency and ensuring a uniform electric field while functioning on both terminal sides. This advantage allows for the innovative design of a large-area device featuring a double EC cathode structure on both sides and a central zinc mesh anode. Utilizing this configuration, combined with the NIR transmittance modulation of WO ∙ 2HO, we implemented a four-mode functioning EESD that can freely control visible and NIR light (Fig. 4a). By separately managing the coloration of WO ∙ 2HO for NIR light and polyviologen for visible light, four smart window operation modes were realized: transparent, visible colored, NIR colored, and fully colored. The transmittance spectra from 320 to 2500 nm for the four modes are shown in Fig. 4b. The device in transparent mode demonstrates high transmittance in the visible and NIR bands from 380 to 2500 nm. In the visible colored mode, the device displays intense coloration in the 380 to 1000 nm range while maintaining high transmittance from 1000 to 2500 nm. In contrast, the NIR colored mode shows the opposite transmittance patterns, and the fully colored mode exhibits low transmittance throughout the 320 to 2500 nm range. At representative wavelengths of 600 nm (visible) and 1800 nm (NIR), the device demonstrated transmittance contrasts of 57.53% and 43.76%. All four models exhibit low transmittance in the ultraviolet (UV) region from 320 to 380 nm, blocking UV rays from reaching the human body. Actual photos of the four modes are shown in Fig. 4c (2.5 × 2.5 cm²), and large-area device fabrication has also been demonstrated (10 × 10 cm, Figure S4, Supplementary Information).

The integrated optical transmittance (T) and solar irradiance-weighted transmittance (T') of the polyviologen|Zn mesh|WO₃ ∙ 2H₂O EESD were calculated for the visible (380-780 nm), NIR (780-2500 nm), and entire solar spectrum (380-2500 nm) regions. The results are presented in Fig. 4d and Table S2 (Supplementary Information). It is evident that the visible colored and NIR colored modes precisely regulate the transmission or blocking of solar light and radiation within respective spectral ranges. Meanwhile, the transparent and fully colored modes regulate solar radiation transmittance across the spectrum, with 54.75% and 2.34%, respectively, effectively controlling full-spectrum light transmittance.

The polyviologen and WO ∙ 2HO layers are connected to the Zn mesh via two separate electrochemical circuits. When the device is switched to the colored state by selectively linking the Zn mesh to each cathode, redox reactions occur independently within each circuit. Specifically, a reversible redox reaction takes place in the polyviologen|Zn circuit, while the WO ∙ 2HO | Zn circuit simultaneously undergoes the following reaction sequence:

Each cathode operates through its redox mechanism without interference. This is further supported by the nearly unchanged EC performance observed in subsequent tests.

The response times of the polyviologen|Zn mesh|WO ∙ 2HO EESD from transparent to fully colored mode were evaluated at 600 nm and 1800 nm, representing the visible and NIR regions, respectively (Fig. 4e). Since polyviologen and WO ∙ 2HO undergo transmittance changes in respective spectral ranges, the switching at 600 nm and 1800 nm reflects the behavior of polyviologen and WO ∙ 2HO -based Zn mesh EESDs, respectively. The results show an ultrafast bleaching time of 0.8 s and a slightly delayed coloration time of 1.8 s at 600 nm, while at 1800 nm, the switching times are 3.2 s and 1.6 s, substantially outperforming most tungsten oxide-based devices. Additionally, the coloration efficiency at 600 nm and 1800 nm was obtained, yielding values of 296.28 and 233.61 C cm, respectively (Fig. 4f). The driving dual cathode coloration increases the charge density, which causes a tweak in CE values; however, these values remain highly competitive compared to most previous studies. The two independent circuits exhibit long-term cycling stability, operating continuously for 6,000 s with negligible performance degradation (Fig. S5, Supplementary Information). Both the polyviologen|Zn and WO ∙ 2HO | Zn circuits demonstrated a bistability capability, maintaining their colored states for approximately 30 min after disconnection (Fig. S6, Supplementary Information). This behavior is beneficial for enhancing the energy-saving potential of the smart window.

Energy consumption and energy return are key performance indicators for evaluating EESDs, playing a crucial role in energy conservation and emission reduction. The energy required to fully charge the polyviologen and WO ∙ 2HO cathodes with the Zn mesh anode and the energy that can be fully discharged were estimated. The percentage ratio between these values represents the energy return efficiency. As shown in Fig. 5a, a fully discharged polyviologen|Zn mesh circuit was initially subjected to a potentiostatic charging process at -2.0 V for 60 s (bleaching process), during which a total energy input of 132.50 mWh m was recorded (Fig. S7a, Supplementary Information). The device was then galvanostatic discharged under 0.3 mA cm (coloring process), releasing a total energy output of 58.89 mWh m (Fig. S7b, Supplementary Information). Finally, an additional potentiostatic discharge at 0 V for 60 s was carried out to complete the final coloring process, consuming a negligible amount of energy, approximately 0.0018 mWh m (Fig. S7c, Supplementary Information). Overall, the charge-discharge cycle consumed 132.5018 mWh m and delivered an energy output of 58.89 mWh m, yielding an energy return efficiency of 44.4%. Similarly, the energy return efficiency of the WO ∙ 2HO | Zn mesh circuit was determined to be 31.0% (Fig. 5b and S7d-f, Supplementary Information). The recovered energy can be used to power other electrical devices, such as lighting or heating/cooling systems. The polyviologen|Zn mesh and WO ∙ 2HO | Zn mesh circuits produce open-circuit voltages (OCV) of 1.23 V and 1.05 V, respectively. When two identical devices are connected in series, an OCV greater than 2 V can be achieved, sufficient to power an LED light (1.8 V, 2.2 W, Fig. 5c and S10, Supplementary Information).

Notably, both circuits are capable of self-coloration and self-bleaching. By directly connecting the Zn mesh anode to either the polyviologen cathode or the WO ∙ 2HO cathode, the internal potential difference drives spontaneous redox reactions, completing the coloration within 5 s (Fig. 5d). After disconnecting the electrodes, the device gradually undergoes complete self-bleaching within around 30 min. This cycle operates under zero energy consumption, further enhancing the device's energy-saving capability and contributing to the realization of net-zero energy buildings.

We analyzed the solar irradiance spectra of our device under different operating modes and calculated the corresponding Solar Heat Gain Coefficient (SHGC) values using Equations S1-S3 (Supplementary Information). The resulting SHGC values are as follows: fully transparent - 0.654, NIR colored - 0.564, visible colored - 0.348, and fully colored - 0.253. These findings highlight the device's strong ability to regulate solar radiation (Fig. 6a). Thanks to its dual-cathode configuration, the device supports bidirectional use based on weather demands (Fig. 6b). For example, under hot climates, the electrode with higher photothermal conversion efficiency can be oriented toward the exterior. The thermal conductivity of the surrounding materials significantly influences heat transfer, as indicated by Eqs. S4 and S5 in the Supplementary Information, which show that thinner layers facilitate faster thermal conduction. Accordingly, a 0.15 mm PET layer is placed on the exterior side to promote heat dissipation, while a thicker 0.6 mm PET layer is used on the interior side to minimize heat influx. Conversely, in cold climates, the device orientation can be reversed to enhance indoor heat retention. To quantify the photothermal conversion performance, we conducted laser-based photothermal measurements for both polyviologen and WO·2HO (Fig. 6c, d), yielding conversion efficiencies of 35.8% and 62.3%, respectively (Eqs. S6-S9, Tables S3 and S4). These distinct conversion efficiencies enable the device to perform optimally in diverse thermal environments, thereby broadening its usability across various climate zones.

To evaluate the potential of four-mode polyviologen|Zn mesh|WO ∙ 2HO EESDs as smart windows for buildings -- capable of controlling dual-band light transmittance, regulating indoor temperature, and enhancing energy efficiency -- we conducted a 24-hour outdoor real-world model test. As shown in Fig. 6e, we constructed a house model consisting of five identical rooms, each measuring 14 (L) × 14 (W) × 20 (H) cm. Each room had an identical upward-facing window equipped with a 10 × 10 cm sized EESD in four different modes: transparent, visible colored, NIR colored, and fully colored, along with a PET film for comparison. The house model was positioned on the rooftop of an unobstructed building in Busan, South Korea, with the windows facing upward. A temperature logger was placed inside each room to record temperature data throughout the day on October 1, 2024 (sunny). The recorded minute-by-minute temperature variations in each room exhibited significant differences between 8:00 and 18:00 due to variations in solar radiation intensity and the window's ability to block radiation (Fig. 6f). At solar noon, around 12:34, the interior temperatures of the model rooms with PET film, transparent, visible colored, NIR colored, and fully colored modes as windows reached 71.3 °C, 70.0 °C, 53.7 °C, 66.4 °C, and 46.3 °C, respectively, while the outdoor ambient temperature (with walls fully shaded) was 37.9 °C. Figure 6g illustrates the temperature difference between the room and ambient temperatures using PET film and fully colored mode as the window at various local times. In the lower part of this figure, we also indicate the recommended usage times for the four modes. During midday, when solar radiation is intense, the fully colored mode helps lower the temperature and block the glaring sunlight. During dawn and dusk, the NIR-colored mode enhances brightness. At other times, the self-bleaching transparent mode can be activated to store energy efficiently. Additionally, visible light blocking can be applied at any time to ensure privacy as needed.

Along with the 24-hour outdoor real-world test, we also performed EnergyPlus building HVAC energy consumption simulations to assess the device's applicability for each month of the year, both in a specific city and across all U.S. states over the course of an entire year. We created a 40% window-to-wall ratio office model with typical office lighting, HVAC, and occupancy parameters (Table S5 and Fig. S9, Supplementary Information). The HVAC energy consumption for Phoenix, Arizona, was simulated monthly using climate data from the EnergyPlus website. Phoenix has distinct seasonal temperature variations, serving as a representative example to illustrate seasonal energy efficiency changes. The simulations compared commercial low-E glass (SunGuard® SNX 62/27 on Clear) and the polyviologen|Zn mesh|WO ⋅ 2HO EESD as the sole variable window material, with the smart window set to operate in its most energy-efficient mode at various times and temperatures. The related data is presented in Fig. 7a and Table S6 (Supplementary Information). It was observed that the polyviologen|Zn mesh|WO ⋅ 2HO EESD, in comparison to low-E glass, proved to be more energy-efficient throughout the year, achieving energy savings of 10%-15% during the winter months and saving 10-18 MJ m during the summer months. Expanding the use of this smart window to other U.S. states, we simulated the annual building HVAC energy consumption in 16 cities representing typical climates across the country. The findings indicated comparable energy savings, with a 40 MJ m reduction accounting for more than 10% of the total annual HVAC energy consumption (Fig. 7b). Similar outcomes were obtained when implemented across the 50 U.S. states in 15 different climate zones, demonstrating the widespread feasibility of the smart window's energy-saving capabilities (Fig. 7c). The U.S. energy-saving map revealed a clear trend showing that the energy-saving effect is more significant in warmer regions at lower latitudes. This is attributed to the ability of the polyviologen|Zn mesh|WO ∙ 2HO EESD to block both light and thermal radiation.

Our smart windows provide substantial economic and environmental benefits, which can be clearly seen in the costs and carbon emissions related to the energy savings. As shown in Table S7 (Supplementary Information), the average energy savings across U.S. states is calculated to be 26.72 MJ m. The U.S. Energy Information Administration (EIA) reported that the total commercial building area in the U.S. was about 97 billion square feet in 2018, allowing for a calculation of total energy savings across the country to be as high as 66.87 billion MWh. We have also obtained the levelized cost of electricity (LCOE) for different energy generation methods. The economic costs of producing this large amount of electricity using nuclear, coal, gas, photovoltaic, onshore, and offshore wind power generation systems are estimated to range from $2.67 trillion to $7.35 trillion (Fig. 7d and Table S8, Eq. S10, Supplementary Information). By examining the carbon emissions per unit of electricity generated by various technologies, the carbon emissions associated with these energy savings can be calculated (Eq. S11, Supplementary Information). Even clean energy sources, like nuclear, photovoltaic, and wind, produce 0.87 billion tons of CO, while coal, a highly polluting energy source, has an astounding 66.94 billion tons of CO emissions. These results, gathered from tests and simulations at various scales, demonstrate that this dual-electrode, four-mode smart window effectively reduces building HVAC energy consumption. It holds significant economic and environmental value, aiding in energy conservation, emission reduction, and achieving net-zero energy buildings globally.