2.1 Effects of Alkali Doping on Kesterite Absorber Quality and Device Performance

The solution-processed Cu₂ZnSnS₄ (CZTS) solar cells were fabricated through three key steps: (1) precursor preparation: copper(I) chloride (CuCl), zinc acetate [Zn(OAc)₂], and tin(IV) chloride pentahydrate (SnCl₄·5 H₂O) were dissolved in N,N-dimethylformamide to create the precursor solution; (2) spin-coating and annealing: the solution was spin-coated onto substrates and then annealed to form solid precursor films; and (3) chalcogenide atmosphere annealing: the films were annealed in a chalcogenide-rich atmosphere to promote crystallization and grain growth.^{[34, 35]} Detailed methodologies are provided in the Experimental Section. To control Na diffusion, soda–lime glass (SLG) substrates with a SiO_x barrier layer were used. Lithium perchlorate (LiClO₄) and sodium perchlorate (NaClO₄) were selected as alkali sources due to their solubility and thermal stability, as listed in Table S1, Supporting Information. High concentrations (0.5 m) of these alkalis were spin-coated onto the precursor films as top layers, as schematically illustrated in Figure 1a, thereby avoiding potential alterations of the precursor solution chemistry that could occur with alkali introduction into the solution. By introducing Na and Li, we systematically assessed their individual and combined impacts on device performance.

Figure 1b depicts the configuration of the CZTS solar cell, comprising a glass/Mo/CZTS/CdS/ITO structure. The representative current density–voltage (J–V) curves of the solar cells, shown in Figure 1c, reveal the substantial impact of different alkali doping strategies on device performance under standard AM1.5G illumination. Statistical box plots of the key PV parameters V_OC, short-circuit current density (J_SC), FF, and PCE are presented in Figure 1d, further highlighting the distinct improvements brought about by Na and Na–Li codoping. The reference (Ref.) device exhibits the poorest performance across all PV metrics, with a PCE of ≈6%. Upon Li doping, the device performance shows only modest improvement, indicating that Li doping alone does not significantly enhance CZTS performance. In contrast, Na doping leads to substantial improvements across all PV metrics, particularly in FF and PCE. This suggests that Na introduction significantly enhances charge carrier transport and reduces device resistance, likely due to improved crystallinity and larger grain size, which will be discussed in a later section. The most notable performance boost is observed with Na–Li codoping. The codoped device achieves the highest values for V_OC (722 mV), J_SC (23.4 mA cm⁻²), FF (59.3%), and PCE (10.1%). Specifically, the Na–Li codoped CZTS device demonstrates an increase in V_OC due to the increase the carrier concentration, reduced nonradiative recombination, and potential grain boundary passivation, which will be discussed in detail in the following section. The FF is improved as a result of reduced series resistance and enhanced charge carrier transport, benefiting from the synergistic effects of both Na and Li. Consequently, the PCE is elevated from 6.4% in the Ref. device to 10.1% in the Na–Li codoped device, representing a significant overall efficiency enhancement.

The external quantum efficiency (EQE) spectra (Figure 1e) provide further insights into the impact of alkali doping on the optical and electronic properties of the CZTS devices. The Ref. device exhibits a broad EQE response peaking between 500 and 700 nm, corresponding to an E_g of ≈1.55 eV, but with lower overall EQE values compared to the alkali-doped devices, indicating higher recombination losses and less efficient photon–electron conversion. Li doping results in a modest improvement in EQE, suggesting slight enhancement in charge collection efficiency. In contrast, Na doping significantly improves the EQE across the 500–900 nm range, indicating reduced recombination losses. The slight reduction in bandgap to 1.53 eV further supports enhanced light harvesting, particularly in the near-infrared region. The Na–Li codoped device achieves the highest EQE across the spectrum, demonstrating that the codoping strategy minimizes recombination and maximizes charge collection by improving absorber quality, resulting in superior overall device performance. The Urbach energy (E_U) was derived from the EQE data (Figure S1, Supporting Information). The Ref., Li-doped, and Na-doped samples exhibit similar E_U values around 38 meV, indicating comparable levels of band tailing. However, the Na–Li codoped sample shows a lower E_U of 34 meV, suggesting reduced band tail states associated with defects such as Cu_Zn and [2Cu_Zn + Sn_Zn].^{[36, 37]} This reduction suggests that Li helps suppress detrimental defects, which directly contributes to the improved V_OC and PCE of the device. The lower E_U in the Na–Li codoped sample highlights the effectiveness of codoping in enhancing absorber quality and overall device performance.

To understand the mechanisms underlying the performance enhancements observed with different alkali doping strategies, we conducted detailed analyses of the films before and after sulfurization using X-ray diffraction (XRD), Raman spectroscopy, and scanning electron microscopy (SEM). The XRD patterns of the precursor films (Figure 2a) display diffraction peaks which correspond to the (211), (204), and (312) planes of the kesterite phase. Their broad width and weak intensity are indicative of a low-crystallinity kesterite structure. Notably, no secondary phases are observed in the precursor films, suggesting a pure kesterite composition free from impurities. Additionally, no peaks corresponding to NaClO₄ or LiClO₄ are detected, implying that these compounds, if present, are likely amorphous and reside on the film surface. Raman spectroscopy (Figure 2d) complements the XRD findings by confirming the presence of low-crystallinity kesterite phases in the precursor films. In the Na–Li codoped films, vibrational peaks attributed to (ClO₄)⁻¹ at around 930 cm⁻¹ are detected, indicating the presence of these compounds on the film surface. This suggests that the alkali salts remain on the surface in an amorphous state after spin-coating and before sulfurization. Additionally, the presence of LiClO₄ and NaClO₄ on the film surface is further confirmed by SEM image shown in Figure S2, Supporting Information.

It has been previously reported that low-crystallinity kesterite films undergo a direct, single-phase transformation from an amorphous kesterite structure to a highly crystalline absorber layer during heat treatment in a chalcogenide atmosphere (S or Se).^{[35, 36, 38]} This transformation pathway avoids the formation of secondary phases and is critical for preparing high-quality kesterite films. Our findings are consistent with this mechanism. After sulfurization, the XRD patterns of the absorber layer films (Figure 2b) reveal a significant improvement in crystallinity. The characteristic peaks of kesterite become sharper and more intense, indicating larger grain sizes and enhanced crystal ordering. Notably, the Na-doped and Na–Li codoped films exhibit stronger diffraction peaks with narrower full width at half maximum (FWHM) values, as shown in the zoomed (211) peak (Figure 2c). This observation suggests that Na plays a crucial role in promoting grain growth and improving crystallinity. In contrast, the Li-doped films do not show significant enhancement in crystallinity; the diffraction peaks remain relatively broad and weak compared to the Na-doped and Na–Li codoped films. As mentioned above, this aligns with previous studies indicating that Na facilitates the formation of a Na₂S_x liquid phase during sulfurization, which acts as a flux to promote grain growth and improve crystal quality.^[39] Raman spectroscopy of the absorber films (Figure 2e) further corroborates the XRD results. The Raman spectra reveal distinct peaks at 285 and 330 cm⁻¹, associated with kesterite CZTS, with no detectable secondary phases such as ZnS or SnS. The absence of secondary phases in both XRD and Raman analyses emphasizes the effectiveness of Na doping in enhancing crystallinity without introducing unwanted compounds.

SEM images presented in Figure 3 further illustrate the significant impact of alkali doping on the grain morphology of CZTS thin films. The Ref. film exhibits small, irregular grains with numerous grain boundaries, likely serving as recombination centers that limit device performance. This observation aligns with the poor crystallinity indicated by the broad and weak peaks in the XRD patterns and correlates with the low-efficiency metrics of the Ref. devices. The Li-doped film shows only minor improvements in grain size and morphology compared to the Ref. The SEM images and XRD patterns of the Li-doped film remain similar to those of the Ref., suggesting that Li alone does not significantly improve crystallinity or promote grain growth. In contrast, the Na-doped film displays much larger, well-defined grains with reduced grain boundary density, as evidenced by the SEM images. This improved grain morphology corresponds with sharper and more intense peaks in the XRD patterns, indicating enhanced crystallinity and larger grain sizes. The larger grains and fewer grain boundaries lead to decreased recombination and improved charge transport, resulting in significant enhancements in device parameters such as J_SC, FF, and PCE. The Na–Li codoped film exhibits grain morphology similar to that of the Na-doped film, with no significant further improvement in grain size or uniformity observed in the SEM images. Despite the similar crystal morphology, the Na–Li codoped films demonstrate better optoelectronic properties compared to the Na-doped films.

To analyze the depth distribution of elements within the films and understand the effects of Na and Li incorporation, we employed glow discharge optical emission spectroscopy (GDOES). The results, shown in Figure S3a–e, Supporting Information, indicate that the primary kesterite elements—Cu, Zn, Sn, and S—are uniformly distributed throughout the depth in each sample. A clear Na signal is present in the Ref. sample, suggesting that despite the application of a SiO_x layer beneath the Mo as a Na diffusion barrier (Figure 1b), Na migration from the substrate into the CZTS layer still occurs at the high annealing temperature of 640 °C. However, this substrate-derived Na did not significantly enhance film morphology, likely due to its low concentration. So, Na diffusing from the SLG does not dominate the absorber quality enhancement. In the Na-doped samples (Na-doped and Na–Li codoped), the Na signal is significantly stronger, indicating successful incorporation of Na from the spin-coated NaClO₄ layer. Interestingly, in the Li-doped samples (Li-doped and Na–Li codoped), almost no Li is detected, while the Na content is notably higher than in the samples without Li (Figure S3e, Supporting Information). GDOES analysis of the precursor film (Figure S3f, Supporting Information) confirms a substantial Li presence on the film surface before sulfurization, indicating that Li is not lost during the spin-coating process.

Yang et al. previously reported on the competitive exchange behavior between Na and Li in kesterite films, where Na tends to repel Li from the film and their interaction promotes Na diffusion.^[40] This phenomenon explains the consistent observation of significant Li loss during experiments. Our study aligns with this mechanism, showing that Na repels Li during the high-temperature sulfurization process, leading to severe Li depletion and its undetectability via GDOES in the final films. This Na–Li exchange mechanism results in Li effectively enhancing the Na content within the film, as illustrated in Figure S3e, Supporting Information.

Photoluminescence (PL) and time-resolved PL (TRPL) measurements were conducted to elucidate the role of alkali doping in enhancing the optoelectronic properties of CZTS devices (Figure 4). The Ref. sample exhibits the weakest PL intensity and the shortest carrier lifetime (≈2 ns), indicating a high density of recombination centers. This high recombination rate leads to poor charge collection and low device performance, consistent with the inferior PV metrics observed. Li doping results in a slight increase in PL intensity, reflecting modest improvements in reducing recombination losses. However, the overall effect on device performance remains limited, aligning with the minimal enhancements observed in the J–V characteristics. In contrast, Na doping significantly enhances PL intensity and extends the carrier lifetime to ≈10 ns. The increased PL intensity and carrier lifetime indicate improved absorber quality and reduced nonradiative recombination, likely due to enhanced crystallinity and fewer grain boundaries facilitated by Na incorporation, which correlates with better device performance. The Na–Li codoped device exhibits the highest PL intensity among all samples studied, indicating a further reduction in recombination. This enhanced PL behavior is attributed to the synergistic effect of Na and Li codoping. The introduction of Li alongside Na likely promotes the formation of beneficial shallow defect states such as Li_Cu, Li_Zn, and V_Cu, which enhance p-type conductivity and help passivate detrimental defects and grain boundaries of absorber, as supported by DFT calculations.^{[28, 33, 41]}

Alkali doping is known to increase the carrier concentration for the kesterite, thereby improving overall device performance. To quantify this effect, capacitance–voltage (C–V) measurements were performed, and the results are presented in Figure S4, Supporting Information. The Ref. and Na-doped devices exhibit similar carrier concentrations, around 2 × 10¹⁶ cm⁻³. However, Na–Li codoping significantly increases the carrier concentration to ≈6 × 10¹⁶ cm⁻³. This substantial increase in carrier concentration can be attributed to two primary factors, as reported in both DFT and experimental literature: 1) formation of shallow acceptors: the incorporation of Li into the film leads to the formation of shallow acceptor defects such as Li_Cu and Li_Zn,^{[21, 26, 34]} enhancing p-type conductivity; 2) assisting Na incorporation: Li facilitates a higher Na content within the film, as observed in the GDOES analysis. The increased Na content promotes the formation of beneficial copper vacancies (V_Cu), further improving carrier density.^[42]

2.2 Efficient CZTS Solar Cells for IPV Applications

As mentioned before, for optimal performance in indoor conditions, the IPV device needs to meet two key criteria: (1) a bandgap of 1.7–2.0 eV to match the typical indoor light spectra,^[21] and (2) high shunt resistance to minimize leakage currents, which become the dominant source of carrier losses under low-light conditions.^[43] In this study, the champion CZTS device features an absorber layer with a bandgap of 1.55 eV. Although this is somewhat lower than the ideal value for IPVs, it demonstrates the potential for further performance improvement through bandgap tuning.

Our results demonstrate that high-efficiency CZTS solar cells exhibit remarkable adaptability and efficiency under varying lighting conditions, including both simulated sunlight (AM1.5G) and artificial indoor lighting. Figure 5a shows the J–V curve of the champion CZTS device under AM1.5G illumination and dark conditions. The device displays a high shunt resistance (≈10 000 Ω cm²), which is ideal for indoor PV applications. The device's response to varying light intensities under the AM1.5G spectrum is shown in Figure 5b. As the light intensity decreases from 1 to 0.1 sun, the FF improves from 59.1 to 67.1%, with only a slight reduction in V_OC. The stability of V_OC across these intensities indicates that the device effectively minimizes carrier recombination losses. This demonstrates that CZTS solar cells efficiently harvest energy under low-light conditions, such as scattered sunlight commonly found indoors during the day.

Figure 5c provides spectrum for 20 different indoor lighting conditions used to characterize device performance in this study, including five color temperature lights (2700, 3000, 4000, 5000, and 6000 K) and four intensity levels for each temperature light. We used this wide range of spectra for performance testing, rather than a single-color temperature and illumination intensity, to accurately evaluate the real-world performance of solar cells under complex lighting conditions and to simulate the device's operation in realistic indoor environments. Figure 5d presents the J–V curves under the various indoor lighting conditions depicted in Figure 5c. The device maintains high FF values under most of these conditions, and although V_OC shows a slight drop compared to 1 sun illumination, the overall efficiency remains high; the PV parameters are detailed in Table 1 and Table S2, Supporting Information. Under specific conditions, particularly those with high intensities near the light source and warm color temperatures (e.g., 2700 and 3000 K), the CZTS device achieves a PCE of up to over 15.0%, highlighting its potential for indoor PV applications. The slight decrease in V_OC under low-light conditions is consistent with the expected influence of light intensity on V_OC; however, the higher FF values help to maintain a high PCE overall.

Figure 6 displays filled contour plots for V_OC, J_SC, FF, and PCE across the range of tested indoor lighting conditions. These plots illustrate the strong performance of CZTS solar cells under most conditions. As shown in Figure 6a, the device maintains a stable V_OC across various color temperatures and intensities, indicating effective charge separation and low recombination rates. The J_SC plot in Figure 6b reflects the device's capability to maintain decent current generation even as lighting conditions vary. In terms of FF (Figure 6c), the device consistently achieves values above 65% under moderate to high light intensities across all color temperatures, confirming the beneficial effects of high shunt resistance. This behavior contrasts sharply with previous reports on kesterite solar cell under low-light condition, including CZTSe, CZTS, and CZTSSe, which exhibited significant drops in V_OC and FF compared to AM1.5G illumination, as summarized in Table S2, Supporting Information. This resilience against changes in lighting conditions makes CZTS solar cells particularly well suited for indoor applications, where they may be exposed to a variety of artificial and natural light sources. Finally, the PCE plot (Figure 6d) reveals that the CZTS device achieves efficiencies exceeding 14.5% under most lighting conditions, with peak performance around 15.0% under 3000 K at 100% intensity (2.93 mW cm⁻², 9586 lux), closely resembling conditions near the light source. This level of efficiency is promising for indoor PV applications aimed at powering IoT devices.

This comprehensive testing illustrates the versatility of CZTS solar cells in adapting to various light sources typically found indoors, such as LEDs, fluorescent lamps, halogen lamps, and even indirect sunlight. Notably, the performance remains robust across all conditions, underscoring the potential of CZTS solar cells for indoor applications where light sources can vary significantly in intensity and spectral distribution, such as rooms that receive both artificial light and scattered sunlight.

Building on the promising efficiency of CZTS solar cells under indoor lighting, we evaluated their practical application in powering commercial IoT devices—device 1: Philips Hue Motion Sensor, device 2: Pressac CO₂ Temperature & Humidity Sensor, and device 3: Honeywell HIH6130 Temperature & Humidity Sensor—each with varying power demands of 0.02, 0.12, and 0.28 mW cm⁻², respectively, under indoor illumination. Detailed simulations regarding the powering of devices 1–3 are summarized in the Supplementary Information. Figure 7 illustrates the power generation density (PGD) of the CZTS solar cell under 3000 K lighting and its comparison with the power consumption density (PCD) of the IoT devices. The shaded yellow area represents the expected indoor light intensity range, between 0.3 and 3 mW cm⁻², which is typical for artificial lighting in indoor environments.

In this simulation, the PGD line intersects with the PCD requirements of each IoT device, revealing the minimum light intensities needed for each device to function. For IoT Device 1, which requires 0.02 mW cm⁻² to operate, the PGD exceeds the PCD across the entire indoor light intensity range. This indicates that the CZTS solar cell can effectively power this device even under very low-illumination conditions, with sufficient energy to meet the device's operational needs and potentially store excess energy. IoT devices with higher power demands, such as device 2 (0.12 mW cm⁻²) and device 3 (0.28 mW cm⁻²), require more intense illumination to reach the power balance point where PGD equals or surpasses PCD. For device 2, this threshold is ≈0.8 mW cm⁻², while for device 3, it is around 1.85 mW cm⁻². Thus, device 2 can operate in moderately low indoor lighting, whereas device 3 requires more intense lighting—such as direct exposure to indoor LEDs—to function reliably.

These results demonstrate that the high efficiency of CZTS solar cells in indoor settings enables them to produce a sufficient amount of electrical energy, meeting the operational power requirements of various IoT devices. This adaptability to diverse indoor lighting conditions confirms CZTS solar cells as a promising technology for sustainable, self-powered IoT systems. Furthermore, the analysis provides a clear framework for determining the conditions under which CZTS solar cells can power specific devices, emphasizing their potential to drive low-power IoT networks in controlled environments like smart homes and offices.