Trapping lead in perovskite solar modules with abundant and low-cost cation-exchange resins

Abstract

One major concern for the commercialization of perovskite photovoltaic technology is the toxicity of lead from the water-soluble lead halide perovskites that can contaminate the environment. Here, we report an abundant, low-cost and chemically robust cation-exchange resin (CER)-based method that can prevent lead leakage from damaged perovskite solar modules under severe weather conditions. CERs exhibit both high adsorption capacity and high adsorption rate of lead in water due to the high binding energy with lead ions in the mesoporous structure. Integrating CERs with carbon electrodes and layering them on the glass surface of modules has a negligible detrimental effect on device efficiency while reducing lead leakage from perovskite mini-modules by 62-fold to 14.3 ppb in water. The simulated lead leakage from damaged large-area perovskite solar panels treated with CERs can be further reduced to below 7.0 ppb even in the worst-case scenario that every sub-module is damaged.

Main

Power conversion efficiencies (PCEs) have exceeded 25% for single-junction perovskite solar cells (PSCs) and 29% for perovskite/silicon tandem solar cells¹. Large-area perovskite modules have been shown to be manufacturable using a scalable coating process, producing efficiencies comparable to those of silicon modules². The stability of PSCs has also been improved significantly, passing most industry standard tests^3,4. All these developments have demonstrated the great potential of perovskite photovoltaics as the next-generation low-cost solar technology. Nevertheless, one key outstanding issue for the adoption of perovskite photovoltaic technology in the real world is the toxicity of lead in the lead halide perovskite light absorbers^5,6. Although there have been extensive endeavours to replace lead in PSCs, all current lead-free PSCs suffer from either much poorer stability, such as the tin-based PSCs, or much lower PCEs, such as double-perovskite-based solar cells, compared with their lead-based counterparts^7,8. Lead is thus still an essential component in perovskites to realize PSCs with both high efficiency and decent operational stability^9,10. Encapsulation strategies have been developed to prevent the leakage of lead in operational cells and modules¹¹. However, perovskite solar modules can be damaged or even broken under severe weather conditions, including storm and hail. Once damaged, toxic lead can leak out of the perovskite layers during rainfall and then enter and contaminate soil and underground water.

Due to the difficulties in eliminating lead from high-performance PSCs, one alternative strategy to minimize the impact of toxic lead on the environment is to prevent lead leakage when perovskite solar modules are damaged or broken. For instance, a self-healing polymer-based encapsulation method has been reported to suppress lead leakage¹¹. The self-healing effect of a polymer encapsulant at higher temperature can prevent water from penetrating into the damaged perovskite solar modules and thus reduce the lead leakage rate by 375-fold compared with that based on ultraviolet-cured resins. Nevertheless, advanced encapsulation techniques cannot completely prevent lead leakage, because lead can still escape the solar panels in rainwater under extreme weather conditions. The hole transport polymer alkoxy-PTEG has also been reported to chelate lead ions and thus reduce lead leakage¹². However, the thickness of the alkoxy-PTEG layer was merely 27 nm due to its limited conductivity, which is not enough to completely adsorb the lead ions in a 500-nm-thick perovskite layer. Recently, lead-adsorbing layers were applied onto both sides of PSCs, and such a chemical approach substantially reduced the lead leakage by 96% (ref. ¹³). Despite the outstanding lead-adsorbing effect, the high cost of the lead-adsorbing material might prevent its application in low-cost perovskite photovoltaics. To effectively trap lead inside damaged perovskite solar modules, it is critical to find a lead adsorbent that not only effectively prevents lead leakage under extreme weather conditions, but also is low-cost and easily integrated into the manufacturing of perovskite solar modules.

Here, we report a method based on lead-adsorbing coating layers that can effectively minimize lead leakage from damaged perovskite solar modules. This lead-adsorbing layer was prepared from cation-exchange resins (CERs) that have both strong bonding and excellent selectivity towards Pb²⁺ cations in water^14,15. CERs are low-cost, chemically robust, water-insoluble and easily applied on both sides as well as on the electrodes of perovskite solar modules. All these features make CERs nearly ideal candidates to prevent lead leakage from damaged perovskite solar modules. Simulated lead leakage tests show that CERs coated onto the surfaces of metal electrode-based solar mini-modules can effectively reduce lead leakage by ~90% compared with the mini-modules without CER layers. Most importantly, we have successfully integrated the lead-adsorbing CERs into carbon electrodes of PSCs without impacting device efficiency, further reducing lead leakage to below the safe level of drinking water according to the US Federal 40 CFR 141 regulation.

Lead adsorption properties of CERs

We studied the application of CER material as Pb²⁺ adsorbent to prevent lead leakage when perovskite solar modules are damaged by severe weather conditions, such as hail and storm, as illustrated in Fig. 1a. We first conducted a brief computational assessment to gain a qualitative idea as to whether and how strong CERs can adsorb and bind Pb²⁺ in aqueous solution. Specifically, we compared the bonding strength of the sulfonate-terminated structural motifs present in CERs with Pb²⁺ and two other common divalent ions, Ca²⁺ and Mg²⁺. In particular, Ca²⁺ has been reported to present in relatively large quantities in rainwater in certain locations¹⁶. Ca²⁺ and Pb²⁺ are reported to have similar binding ability to various chemical species¹⁷, which has been called ‘ion mimicry’^18,19. In the computational assessment, we considered models of sulfonate groups in which each divalent ion is adsorbed by at least two sulfonate groups for charge neutrality. Specifically, we probed two possible models, shown in Fig. 1b: the adsorption of divalent ions by two sulfonate groups that are not connected by the same backbone, referred as ‘di-monomer‘ adsorption, and the adsorption of divalent ions by sulfonate groups attached to the same backbone, referred as ‘dimer’ adsorption. The calculated adsorption energies of Pb²⁺, Ca²⁺ and Mg²⁺ with respect to the vacuum are −20.35, −20.41, −23.69 eV and −21.48, −21.36, −24.43 eV for the di-monomer and dimer cases, respectively. Mg²⁺ is bound more strongly, but this is offset when considering hydration. The reported hydration energies for Pb²⁺, Ca²⁺ and Mg²⁺ are −15.34, −16.50 and −19.92 eV, respectively²⁰, that is, here, the adsorption energies of the three ions relative to the hydration energies are −5.01, −3.91, −3.77 eV and −6.14, −4.86, −4.51 eV, respectively, for the di-monomer and dimer models, respectively. Overall, Ca²⁺ and Pb²⁺ appear to be adsorbed with similar strength in the same microscopic environment when referenced to the vacuum environment, as already observed in past studies of a much broader range of adsorption scenarios^17,18. However, hydration appears to account for the difference between Ca²⁺ and Pb²⁺. When considering the aqueous solution as the reference state, Pb²⁺ is actually adsorbed more strongly than either Ca²⁺ or Mg²⁺. Although qualitative, these exploratory calculations confirm that the CER material is an effective Pb²⁺ adsorbent.

Fig. 1: Lead adsorption properties of CERs.

a, Schematic of how CERs prevent lead leakage due to the strong ionic interaction between Pb²⁺ in decomposed perovskites and sulfonate groups. b, Graphical representations of the computed minimum-energy adsorption model structures for different cations (see text for details). c, Adsorption kinetics of Pb²⁺ by CERs at different temperatures. The curves drawn on top of the data are guides to the eye. The inset shows pseudo-second-order fitting results. The fitting is based on the pseudo-second-order equation: t/q_t = 1/(kq_e)² + t/q_e, wherein k is the pseudo-second order rate constant and q_e and q_t are the amounts of Pb²⁺ adsorbed at equilibrium and after time t, respectively. d, The change in concentration of aqueous solutions of Pb²⁺ of different initial concentration after flowing over 15-cm-long glass covered by CER layers of different thicknesses at 20 °C. The error bars represent the standard deviation for three samples. The inset shows the set-up used for the flowing-water tests in which the aqueous Pb²⁺ solutions were dripped onto the upper end of the glass sheets and then collected at the lower end.

Source data

The specific surface area of the CER powder was measured to be 45.9 m² g⁻¹ with the pore size peaking at 22.6 nm (Supplementary Fig. 1), determined by a Brunauer–Emmett–Teller surface area and pore size analyser. Such a high surface area and large pore size should facilitate the adsorption of lead in water. Then we investigated the intrinsic lead-adsorbing properties of the CER powder. Both kinetic and thermodynamic studies of the lead adsorption properties of the CERs were carried out. As shown in Fig. 1c, the CERs adsorbed most of the Pb²⁺ within 10 minutes regardless of temperature. The adsorption kinetics of Pb²⁺ by the CERs were then fitted by both pseudo-first-order and pseudo-second-order models, as shown in Supplementary Fig. 2a and the inset in Fig. 1c. The results show that the pseudo-second-order model gives a better fitting, which indicates that only the surface of the CERs adsorbs lead and highlights the importance of using a mesoporous CER structure for a large surface area. This agrees with previous studies on the adsorption of other metal ions by CERs²¹. The kinetic constants and amounts of lead adsorbed in the equilibrium state obtained from pseudo-second-order fitting are summarized in Supplementary Table 1. The adsorption isotherm for Pb²⁺ on the CERs was also investigated, as shown in Supplementary Fig. 2b, which reveals a maximum sorption capacity of 410 mg g⁻¹ for lead in aqueous solution.

Next, we tested the adsorption rate of lead by a CER film flushed with lead-contaminated flowing water. We prepared CER films on glass sheets by suspending CER powders in isopropyl alcohol assisted by sonication, and then blade-coated the CER precursor solution onto the top of the glass substrates with different thicknesses ranging from 300 to 1,300 nm. Morphological studies showed the CER layers to have good coverage on the glass surfaces (Supplementary Fig. 3). The surface of the CER-covered glass was highly hydrophilic, evidenced by a small water contact angle of ~13° (Supplementary Fig. 4), which should help water penetrate into the porous structure of the CERs and thus facilitate Pb²⁺ adsorption. In all flowing-water tests in this study, the CER-covered glass sheets or modules were tilted at an angle of ~30° with respect to the Earth’s surface, which is the optimal tilt angle of solar modules for regions at 35° latitude (for instance, North Carolina, USA) to achieve the maximum energy yield throughout the year¹¹. To simulate how fast lead can be adsorbed by the CERs on solar modules when it rains, aqueous Pb²⁺ solutions with concentrations of 10 ppm, 1 ppm and 10 ppb were dripped onto the top end of the glass sheets, as illustrated in the inset of Fig. 1d. The aqueous solutions were collected at the bottom end of the glass sheets and analysed by inductively coupled plasma mass spectrometry (ICP-MS). The flow distance and dwell time of the aqueous Pb²⁺ solutions on the glass sheets were 15 cm (the length of our typical perovskite solar mini-modules) and ~4 s, respectively. As shown in Fig. 1d, a 300-nm-thick CER layer coated on glass immediately reduced the lead content in flowing water by ~30%, independent of the initial lead concentration in the solution. Additionally, the reduction of lead in water did not change much when further increasing the thickness of the CER layer, which suggests that the lead adsorption rate should be limited by the diffusion of Pb²⁺ in the flowing water to the surface of the CERs. These results verify the good lead adsorption properties of the CERs, which exhibit both high adsorption capacity and fast adsorption of lead from water. The CER layers coated on glass surfaces still maintained their lead adsorption capability after being light-soaked for 600 hours under simulated 1 Sun light with a strong UV component (Supplementary Fig. 5), indicating the good light stability of the CERs.

We then tested the lead adsorption performance of the CERs on perovskite solar devices by blade-coating a CER precursor solution onto the metal electrodes of PSCs. The PSCs were fabricated with a positive–intrinsic–negative (p–i–n) architecture of indium tin oxide (ITO)/poly(bis(4-phenyl)(2,4,6-trimethylphenyl)amine) (PTAA)/methylammonium lead triiodide (MAPbI₃)/C₆₀/bathocuproine (BCP)/copper. The morphology of the pristine CER-covered MAPbI₃ solar cells was investigated by scanning electron microscopy (SEM). The top-view SEM image in Fig. 2a shows that a compact CER layer was formed on top of the copper electrode with full coverage, and the cross-sectional SEM images in Fig. 2b,c show that the mesoporous CER layer comprises many nanoparticles with an average particle size of ~50 nm. Such mesoporous structures can vastly increase the surface area and adsorb more lead. We also investigated whether such a CER layer has a detrimental effect on solar cell efficiency because it was coated directly onto the copper electrodes. The current density–voltage (J–V) curves of the MAPbI₃ PSC before and after blade-coating the CERs onto the copper electrodes are presented in Supplementary Fig. 6 and show that the device had an almost identical PCE after depositing the CER layer. The direct coating of the lead-adsorbing CER layer on top of the metal electrodes without a protection layer would simplify the fabrication procedure and thus decrease production cost. To investigate whether the coating of CERs on the copper electrodes can accelerate the degradation of the PSCs due to a potential thermal expansion mismatch, we further evaluated the stability of the encapsulated MAPbI₃ PSCs with and without the CERs under temperature cycling tests following the IEC 61215 protocol. The results presented in Supplementary Fig. 7a,b show that the CERs on top of the copper electrodes did not accelerate the degradation of the encapsulated solar cells after temperature cycling tests for over 50 cycles.

Fig. 2: Characterization of CERs on PSCs.

a,b, Top-view (a) and cross-sectional (b) SEM images of the MAPbI₃ device with CER back coating. c, SEM image of the CERs on top of the copper electrode shown in b.

Lead leakage tests on perovskite mini-modules

We also coated a CER layer on top of the metal electrodes of perovskite solar mini-modules and then studied how effectively it can prevent lead leakage in damaged solar mini-modules. MAPbI₃ perovskite solar mini-modules with the structure of ITO/PTAA/MAPbI₃/C₆₀/BCP/metal cathode were fabricated according to our previously reported method². The average area of all the MAPbI₃ solar mini-modules (Fig. 3a) used in this work was 104.1 ± 2.3 cm². To exclude the impact of module-to-module variation on the results of the lead leakage study, the perovskite solar modules were cut into two almost identical small mini-modules, one of which was used as a control and the other was coated with the CERs. The mini-modules were then encapsulated with a 1.1-mm-thick glass using epoxy coating on the edges of the mini-modules. After that, the mini-modules were broken by dropping ice balls onto them to simulate hail damage to the perovskite modules under real operating conditions by following the ASTM E1038 standard of photovoltaic panel testing for hail impact resistance²². The damaged mini-modules showed star-shaped cracks at the sites of impact, and both the ITO-coated glass and encapsulation glass were broken. The damaged area and size of the cracks in the two mini-modules were very similar in each test, as shown in Supplementary Fig. 8.

Fig. 3: Lead sequestration in perovskite solar mini-modules with CER coating layers.

a–f, Typical process for the preparation of damaged perovskite solar mini-modules for lead leakage tests: photographs of a large MAPbI₃ mini-module with a perovskite-covered area of 102.0 cm² (a), the large mini-module cut into two identical small mini-modules (b), the backs of the two small mini-modules, the right one coated with CERs (c), the fronts (d) and backs (e) of encapsulated small mini-modules prepared by coating the edges with epoxy resin, and mechanically damaged small mini-modules (f). g, Water-soaking test results for the damaged small mini-modules without (w/o) and with (w/) the CER coating layer. The curves drawn on top of the data are guides for the eye. h, Water-dripping test results for the damaged small mini-modules w/o and w/ the CER coating layer (orange, control mini-modules w/o a CER layer encapsulated by a glass slide; blue, mini-modules w/ a back CER layer encapsulated by a glass slide; purple, mini-modules w/ both front and back CER layers encapsulated by a PP sheet). A logarithmic scale is used for the lead concentration. The error bars represent the standard deviation for three samples.

Source data

We first tested lead leakage from the damaged solar mini-modules by water-soaking to simulate the scenario when damaged modules are flooded. In this test, each damaged sub-module was placed in a Petri dish (diameter 15 cm), and 200 ml deionized (DI) water at different temperatures (2, 20, 60 and 85 °C) was poured into the dish to fully cover the whole sub-module. The lead concentration in the contaminated water was then measured by ICP-MS every 30 minutes, and the results are shown in Fig. 3g. At temperatures below 60 °C, the concentration of lead leaking out of both sub-modules was relatively low in the first 30 min, which might be caused by the slow permeation of water into the perovskite layer during this stage. After 1 hour, the lead concentration in the contaminated water from the mini-modules without CER coating increased dramatically, suggesting that the electrode and the charge transport layer were no longer able to prevent water permeation (Supplementary Fig. 9). In addition, the lead leakage rate increased with temperature for the mini-modules without the CER layer. In striking contrast, the damaged mini-modules with CER coating exhibited very small lead leakage regardless of temperature. These results clearly demonstrate that the CER layer can effectively prevent lead leakage from damaged perovskite mini-modules.

We then dripped water onto the damaged solar mini-modules without and with CER layers to simulate heavy rainfall right after the hail impact. At least three mini-modules were tested in each water-dripping test. Each damaged solar mini-module was positioned in a funnel with a tilt angle of ~30° with respect to the Earth’s surface. DI water (pH = 7) was dripped onto the mini-modules at a rate of 5 ml h⁻¹ for 1 hour. The spread area of water on the glass was ~1 cm² to cover the damaged area, corresponding to a heavy rainfall of 50 mm h⁻¹. Contaminated water passing through the damaged mini-modules was collected in a centrifuge tube and then analysed by ICP-MS. The measurement set-up is shown in Supplementary Fig. 10. After dripping water for 1 hour, a yellow-coloured product was observed around the star-shaped cracks in both types of mini-modules. The average lead concentration in the contaminated water from three mini-modules without a CER layer was 13.24 ± 0.25 ppm, and this was greatly reduced to 1.92 ± 0.46 ppm in the water from perovskite mini-modules with CER layers (Fig. 3h). To further simulate heavy acidic rainfall right after hail impact, we also dripped acidic water (pH = 4.2) onto the damaged mini-modules. Similarly, the CER coating layer reduced the average lead concentration from 15.50 ± 0.50 to 2.55 ± 0.14 ppm, and a comparable lead sequestration efficiency of 84% was obtained. Moreover, we conducted an outdoor lead leakage test for 1 hour on a rainy day in North Carolina using the set-up shown in Supplementary Fig. 11. The rainfall was 1.6 mm with a pH of ~6, and the water passing through damaged mini-modules was collected following the same procedure. The ICP-MS analysis showed that the CER coating layer reduced the lead concentration in the collected water from 3.68 ± 0.07 to 0.32 ± 0.04 ppm, corresponding to a high lead sequestration efficiency of 91%. We further extended the outdoor lead leakage test to 9 hours under an accumulated rainfall of 16.7 mm. The ICP-MS analysis showed that the CER layer on the metal electrodes further reduced the concentration of lead leaking into rainwater from 432 ± 16 to 33.6 ± 2.9 ppb. It is noted that the lead concentration in the water from the two systems in the 9-hour outdoor test was lower than that in the 1-hour test. This can be explained by the most soluble lead ions leaking out of the module at an early stage, with the following rainwater continuing to dilute the initial rainwater. These results clearly verify that CER layers can effectively reduce lead leakage when perovskite solar modules are damaged.

During the water-dripping experiment, water penetrated into the damaged mini-modules through cracks and then dissolved the perovskite material. In the CER-treated solar modules, most of the dissolved lead is captured by the CER layer coated on the electrodes, but a small portion of lead can still leak out to the glass surface of the damaged modules. To also block this lead leakage pathway, we blade-coated a CER layer onto the glass side of mini-modules to further adsorb leaked lead. A 300-nm-thick CER layer was found to have no notable effect on the transmittance of ITO/glass substrates, nor on the external quantum efficiency of MAPbI₃ solar cells, as shown in Supplementary Fig. 12. We also used a 1-mm-thick polypropylene (PP) sheet to encapsulate the metal electrodes of the mini-modules, because a polymer frame is generally used in solar module encapsulation. Photographs of a PP-encapsulated solar mini-module before and after ice impact are shown in Supplementary Fig. 13. We conducted the same neutral and acidic water-dripping tests and found that the lead leakage rates were further reduced. The concentrations of lead in the water-dripping tests were further reduced to 96.5 ± 8.4 ppb in DI water and 157 ± 17 ppb in acidic water. These results suggest that a thin CER layer on the glass substrate together with polymer encapsulation can further reduce lead leakage.

PSCs with CER-integrated carbon electrodes

We further reduced lead leakage by incorporating CERs into carbon-based electrodes. Carbon materials have shown great promise as electrodes for perovskites due to their low cost, solution processability and excellent stability with perovskites^23,24,25, whereas most known metals used in solar cells react with perovskites^26,27,28,29. Highly stable PSCs have been reported using carbon electrodes^30,31,32,33, and several companies are scaling up this technology. Carbon electrodes made from carbon paste containing both graphite flakes (a few μm in size) and carbon black (50–100 nm in size) are relatively porous. The loose structure can accommodate small-sized CER nanoparticles without sacrificing the conductivity of the carbon electrodes (Fig. 4a). To test this, we mixed carbon paste and CERs in various mass ratios of 10:1, 5:1 and 2:1, and tested the conductance of neat carbon paste and the three carbon/CER films with the lateral structure illustrated in the inset of Fig. 4b. The current–voltage (I–V) curves measured in the absence of light in Fig. 4b show that the conductance decreases with increased loading of CERs. The carbon films with 10 and 20% CERs largely maintain the conductance of the pure carbon electrode, whereas 50% CERs notably decreased the conductance of the carbon electrodes. Solar cells were subsequently fabricated with the structure of ITO/PTAA/Rb_0.05Cs_0.05FA_0.85MA_0.05PbI_2.85Br_0.15/C₆₀/SnO₂/carbon electrode, where MA is methylammonium and FA is formamidinium. The PSCs based on carbon paste with 10 and 20% CERs exhibited PCEs of 18.0 and 17.9%, respectively (Fig. 4c and Table 1), which are comparable to that of the device prepared with neat carbon paste (18.1%). More device characterization data, including device hysteresis, stabilized power output and statistical analysis, can be found in Supplementary Fig. 14. These efficiencies are among the highest observed for PSCs using carbon electrodes. This agrees with the unchanged morphology of the carbon paste after adding 20% CERs, as shown by the cross-sectional SEM images in Fig. 4d, which also show that the addition of 20% CERs does not affect the contact between carbon paste and the underlying perovskite/C₆₀/SnO₂ films. The electron dispersive spectroscopy (EDS) mapping image in Fig. 4e shows that sulfur is uniformly distributed on all the carbon clusters. We note that the sulfur signal is also observed in the perovskite layer, which is caused by overlapping EDS peaks of sulfur and lead; the lead signal from the perovskite film is recognized as sulfur in the EDS mapping image shown in Fig. 4e (ref. ³⁴). We also tested the carbon/CER mixtures as electrodes in hole-conductor-free solar cells with the structure of ITO/SnO₂/MAPbI₃/carbon electrode, and a similar trend was observed (Supplementary Fig. 15 and Supplementary Table 2). These results verify that a modest amount of CERs added to carbon paste does not sacrifice device efficiency. We then performed the same lead leakage test to investigate how effectively carbon/CER electrodes with a mass ratio of 5:1 can reduce lead leakage in damaged hole-conductor-free mini-modules. Mini-modules with carbon electrodes for water-dripping tests were prepared according to the same procedure as used for metal electrode mini-modules (Supplementary Fig. 16). Pb²⁺ was detected and the S 2p peak shifted to a lower binding energy in the X-ray photoelectron spectroscopy (XPS) measurements of the carbon/CER electrodes after the water-dripping tests (Supplementary Fig. 17). Both indicate that Pb²⁺ was adsorbed by the CERs in the water-dripping tests. The acidic water-dripping tests showed that the application of CERs both in the carbon electrode and on the glass surface effectively reduced lead leakage by 98% from 891 ± 39 to 14.3 ± 1.5 ppb. The residual lead concentration in water is below the safe level (15 ppb) of drinking water according to the US Federal 40 CFR 141 regulation. Such a low Pb concentration in water can be attributed to the greater amount of CERs in the carbon mini-modules compared with in the copper-based modules. We also compared the lead adsorption performance of the CERs with previous reports, as summarized in Supplementary Table 3: the CERs exhibit both low cost and high effectiveness in reducing lead leakage from perovskite devices.

Fig. 4: Lead sequestration in PSCs with CER-incorporated carbon electrodes.

a, Device structure of carbon-based PSCs. b, I–V curves of glass/carbon + CER film/Ag lateral devices with different mass ratios of carbon paste and CERs. The inset shows the lateral device used for I–V characterization. c, J–V curves of PSCs based on carbon electrodes with different carbon paste/CER mass ratios. d, Cross-sectional SEM images of PSCs prepared with neat carbon paste (left) and a carbon/CER (5:1) mixture (right). e, EDS mapping image of a carbon PSC with a carbon/CER (5:1) electrode. The red dots in e show the distribution of sulfur.

Table 1 Device parameters of carbon-based solar cells with different carbon electrodes

Lead leakage simulation on solar panels

To investigate how effectively the resins can reduce lead leakage from solar panels under near worst-case scenarios, we further simulated the lead leakage from large-area solar panels containing multiple mini-modules based on the measurements of mini-modules. On large-area solar panels, water should flow the length of the whole panel and thus lead can be further adsorbed by the CERs coated on the glass (Fig. 5a). We performed a simulation (for details, see the Methods section) to evaluate lead leakage in the worst-case scenario, that is, under heavy and acidic rainfall (50 mm h⁻¹, pH = 4.2) with every mini-module damaged. We assumed rainwater is well spread over the panel and contaminated rainwater from each damaged mini-module has a lead concentration of 14.3 ppb. Further adsorption of lead by the CERs on the glass is regarded a diffusion-controlled process, that is, only lead ions that can diffuse to the surface of the CERs can be adsorbed. Thus, we can use the Lagergren equation³⁵ to describe this process: C₀ − C_t = C₀(1 − e^−kt), wherein k is the adsorption constant, t is the dwell time of water on the surface of the solar panel, C₀ is the initial concentration of lead in rainwater and C_t is the lead concentration at dwell time t. We divided the solar panel into 1980 × 990 pixels, each pixel with a size of 0.1 cm × 0.1 cm (denoted as p × p). We let (i, j) denote the pixel in row i and column j, with pixel (1, 1) situated in the top left corner of the panel. Then we calculated the water flow velocity over each pixel and obtained the lead concentration in rainwater flowing from the first to ith row (the thickness of rainwater on each pixel is presented in Supplementary Fig. 18). Subsequently, we calculated the lead concentration on the (i, j) pixel according to $C_{i,j} = \mathop {\sum}\nolimits_{i = 1}^i {C_i/i}$, and mapped the lead concentration distribution over the whole panel, as shown in Fig. 5b. Clearly, the CER layer on the solar panel gradually further purifies the lead-containing rain. The lead concentration in the rainwater reaching the bottom of the solar panel was simulated to be 6.3 ppb, that is, it is effectively reduced compared with the initial lead concentration leaking out of the damaged mini-modules. These simulated results under a near worst-case scenario further verify the effectiveness of low-cost CER coatings in preventing lead leakage from damaged perovskite solar panels.

Fig. 5: Lead leakage simulation on a damaged carbon perovskite solar panel.

a, Damaged carbon perovskite solar panel with a typical size of 198 cm × 99 cm used for the lead leakage simulation. b, Mapping of the lead concentration on the damaged solar panel in the worst-case scenario, that is, under heavy and acidic rainfall (50 mm h^–1, pH = 4.2) with every mini-module damaged.

Conclusions

In summary, we have developed a low-cost CER-based method to prevent lead leakage from damaged perovskite modules. The coating of lead adsorbent on the surface of metal electrode solar modules can effectively reduce lead leakage independent of the temperature. The integration of the CERs into carbon electrodes and layered on the glass surface reduces the lead leakage from carbon-based mini-modules to 14.3 ppb, which is below the safe level of lead in drinking water. Further encouraging results (<7.0 ppb) were obtained from the simulation of large damaged solar panels, even in the worst-case scenario. Instead of exploring lead-free compositions or reducing the amount of lead in perovskite solar devices, our work opens a pathway to reducing the impact of toxic lead by preventing lead leakage from damaged perovskite solar devices and will accelerate the commercialization of perovskite photovoltaic technology.

Methods

Materials

Strong acid cation-exchange resins (Amberlyst 15, hydrogen form), PTAA (number-averaged molecular mass, M_n = 7,000–10,000), BCP, lead(ii) iodide (99.999% trace metals), lead(ii) bromide (99.999% trace metals basis), caesium iodide, rubidium iodide, N,N-dimethylformamide (DMF), dimethylsulfoxide (DMSO), 2-methoxyethanol (2-ME), toluene, chlorobenzene, lead standard solution (1,000 ± 2 ppm) and propylene glycol methyl ether acetate were purchased from Sigma-Aldrich and used without further purification. Tin(iv) oxide (15% in H₂O colloidal dispersion) was purchased from Alfa Aesar. Methylammonium iodide, methylammonium bromide and formamidinium iodide were purchased from GreatCell Solar.

Device fabrication

Patterned ITO glass substrates (1.5 cm × 1.5 cm for solar cells, 13.0 cm × 8.5 cm for solar modules) were first cleaned by ultrasonication with soap, DI water and isopropyl alcohol (IPA), and then treated with UV/ozone for 15 min before use. The MAPbI₃ solar cells and modules used in this work were fabricated following our previously reported procedure². For the carbon-based perovskite solar cells, the hole transport layer PTAA, at a concentration of 2 mg ml⁻¹ dissolved in toluene, was spin-coated onto ITO glass substrates at a speed of 6,000 r.p.m. for 30 s and then annealed at 100 °C for 5 min. The 1.2 M Rb_0.05Cs_0.05FA_0.85MA_0.05PbI_2.85Br_0.15 precursor solution was dissolved in DMF/DMSO (4:1), and then spin-coated onto the PTAA-covered substrates at a spin rate of 2,000 r.p.m. for 2 s and 4,000 r.p.m. for 20 s, and then 130 µl chlorobenzene was dripped onto the substrates at the 17th second after the start of spin-coating. Subsequently, the films were annealed at 110 °C for 10 min, followed by thermally evaporating a 35-nm-thick C₆₀ layer at a rate of 0.2 Å s⁻¹with an Angstrom Nexdep thermal evaporator. A 15-nm-thick SnO₂ layer was prepared on C₆₀ by the low-temperature atomic layer deposition (LT-ALD) method using a Ultratech/Cambridge Nanotech Savannah S200 system under a base pressure of ~0.1 torr. Tetrakis(dimethylamino)tin(iv) (60 °C) and H₂O (25 °C) were used as precursors for SnO₂ fabrication. The dosing sequences t1, t2, t3 and t4, where t1 and t3 refer to the times of precursor- and oxidant-dosing and t2 and t4 refer to the times for purging, were 0.35, 10, 0.5 and 12 s, respectively. The temperature of the chamber was 100 °C. The growth rate of the LT-ALD SnO₂ layer at 105 °C was ~1.6 Å per cycle. The prepared SnO₂ layer was thermally annealed at 100 °C for 10 min in air. The carbon paste (DD-10, New Seaside Science and Trade), mixed with TEGO AddBond (0.5 wt%, Evonik), was ball-milled overnight before use. Then, the carbon paste and CERs were mixed in certain mass ratios of 10:1, 5:1 and 2:1, and then ground to a powder by hand in an agate mortar. Subsequently, 400 μl propylene glycol methyl ether acetate was added to the carbon paste/CER mixtures (300 mg) and further ground for about 10 min. Finally, the mixture was blade-coated onto the top of the samples with a gap of 50 μm, followed by thermal annealing at 100 °C for 10 min in air. For the hole-conductor-free carbon perovskite solar cells, SnO₂ (15% in H₂O colloidal dispersion) was first diluted with DI water and then spin-coated onto ITO glass at a rate of 5,000 r.p.m., followed by annealing at 170 °C for 30 min. MAPbI₃ films were prepared following the same procedure as detailed above. For the carbon solar modules, SnO₂ was blade-coated onto the ITO module substrates at 70 °C and then annealed at 170 °C for 30 min. Afterwards, MAPbI₃ was blade-coated onto SnO₂ and annealed at 100 °C for 10 min. Subsequently, the carbon paste/CER mixture was blade-coated onto the top of perovskite films and immediately annealed at 100 °C for 30 min. P2 and P3 of the carbon mini-modules were cut by a laser scriber and stainless-steel knife, respectively. The area of the solar cells was 8 mm², as determined by shadow masks. Except for the carbon solar cells, which were prepared in a N₂-filled glove box, the other solar cells and mini-modules were fabricated in air.

Coating of strong acid cation-exchange resins

The CER particles (<300 μm) were first mixed with a high-concentration aqueous NaOH solution under vigorous stirring to turn the hydrogen form to the sodium form. After drying in an oven, the sodium form CERs were finely ground in an agate mortar, and then dispersed in IPA by sonication overnight. The surface area and pore diameter were measured by a Quantachrome NOVA 2000e instrument. The Barrett–Joyner–Halenda (BJH) mode was used to calculate the pore size distribution. The CER supernate was swiped linearly by a film applicator and then blade-coated onto the surface of the perovskite solar cells or modules at a rate of 50 mm s⁻¹. The thickness of the CER layer was tuned by the gap between the blade-coater and substrate. The air knife was operated below 20 psi and no thermal annealing was required.

Device characterization

The J–V characteristics of the solar cells were measured using a Xenon-lamp-based solar simulator (Oriel Sol3A, Class AAA Solar Simulator) and the power of the simulated light was calibrated to 100 mW cm⁻² using a silicon reference cell (Newport 91150V-KG5). All devices were measured using a Keithley 2400 source meter with a forward and backward scan rate of 0.1 V s⁻¹ in air at room temperature, and the delay time was 10 ms. There was no preconditioning before measurement. A metal mask with an aperture (1.6 mm × 3.8 mm) aligned with the device area was used for the measurements. SEM images were taken on a FEI Helios 600 Nanolab Dual Beam System operating at 5 kV. The EDS mapping image was obtained with an EDS Oxford instrument (INCA PentaFET-x3). XPS was performed using a Kratos AXIS spectrometer with an Al Kα anode source (1,486.6 eV) operated at 15 kV and a 10⁻⁸ torr chamber pressure. The data were calibrated using the hydrocarbon C 1s peak (284.8 eV) and processed by CasaXPS software³⁶.

Device encapsulation

The PSCs for the temperature cycling tests were encapsulated by a cover glass sealed by epoxy encapsulant at the edges. A climatic chamber (ESPEC BTX-475) was used to control the temperature. The perovskite solar mini-modules were encapsulated by a 1.1-mm-thick glass or 1-mm-thick PP sheet using Gorilla epoxy on the bottom sides. The top sides of the solar modules were not encapsulated.

Computational details

The adsorption energies of cations were computed using density functional theory methods including (1) non-periodic model calculations, (2) the FHI-aims all-electron code, tight settings and scalar relativity incorporated using ‘atomic zero-order regular approximation’ (recommended standard/default), as described in ref. ³⁷, (3) the Perdew–Burke–Ernzerhof³⁸ density functional was used and the many-body dispersion correction of Ambrosetti et al.³⁹ was added in all calculations, and (4) full geometry relaxation of all probed conformers until all force components were below 10⁻³ eV Å^–1 in magnitude. This level of theory is very well established for capturing even subtle molecular conformation energy differences^40,41, and the same type of approach was also used in a previous broad-ranging reference study of divalent ion adsorption energies and adsorption preferences towards amino acids and their dipeptides¹⁷. Three different cations (Pb²⁺, Ca²⁺ and Mg²⁺) were adsorbed at different sulfonate-bearing site models. Each sulfonate group carries a single negative charge. Only ‘para’ locations of the sulfonate side chain were considered. Cross-linked models (that is, two bridged phenyl rings, indicative of a cross-linked polymer) were not considered. Computationally, it is straightforward to assess ion-binding energies to a substrate with respect to a hypothetical reference state of the isolated ion in a vacuum¹⁷. Conversely, regarding real-world behaviour of the CERs, it is interesting to explore the approximate ion-binding strength to the sulfonate groups on the CERs with reference to the ions in an aqueous solution. Thus, a more relevant reference state for the divalent ions would be their hydrated state. Fortunately, the hydration free energies of common ions with respect to the vacuum state are tabulated in the literature²⁰. Therefore, we can use those hydration energies to connect to divalent ions in a vacuum to obtain an appropriate thermodynamic picture. To bypass the (computationally prohibitive) full conformational complexity of a disordered, cross-linked sulfonated CER bulk polymer, with a possible astronomical number of distinct ion-binding sites and local environments, we instead restricted our analysis to deliberately simple binding scenarios with minimal possible conformational ambiguity, but still indicative of qualitatively appropriate local bonding environments of the ions.

The site models considered were as follows:

In the first adsorption site model (Fig. 1b, top), ions were adsorbed by two styrene-sulfonate monomers, termed ‘di-monomer’. For the divalent cations Pb²⁺, Ca²⁺ and Mg²⁺, the resulting system is neutral (zero overall charge). Three initial di-monomer geometries were sampled for each cation (that is, a total of nine conformers). Note that the conformational degree of freedom is limited by the preferred coordination of each cation, that is, only a single distinct final relaxed geometry resulted for Mg²⁺ and Ca²⁺ from all three different initial geometries, whereas two distinct final relaxed geometries resulted for Pb²⁺. These were energetically separated by 0.15 eV, much smaller than the final qualitative difference found between Ca²⁺ and Pb²⁺ when considering the hydrated ion reference state in the main text of the manuscript.

In the second adsorption site model (Fig. 1b, bottom), ions were adsorbed by a sulfonate-styrene dimer, that is, two side chains connected by a backbone segment that is terminated by methyl groups on either end. One divalent cation (Pb²⁺, Ca²⁺ and Mg²⁺) was considered for the dimer adsorption model, again leading to an overall neutral system in each case. Due to the more limited conformational freedom of this case (assuming that the ion is in direct contact with both sulfonate groups), only one initial dimer model geometry was sampled for each cation.

The corresponding ion adsorption energies, E_ad(ion), with respect to divalent ions in a vacuum can be computed as follows:

$$E_{{\mathrm{ad}}}\left( {{\mathrm{X}}^{2 + }} \right) = E_{{\mathrm{tot}}}\left( {{\mathrm{model}}-{\mathrm{X}}^{2 + }} \right) - E_{{\mathrm{tot}}}\left( {{\mathrm{model}}} \right) - E_{{\mathrm{tot}}}\left( {{\mathrm{X}}^{2 + }} \right),$$

where E_tot(model–X²⁺) denotes the total energy of the sulfonate complex with the X²⁺ ion adsorbed (X = Pb, Ca and Mg), E_tot(model) denotes the total energy of the sulfonate-bearing model molecule(s) without the adsorbed ion and E_tot(X²⁺) denotes the total energy of the isolated ion in a vacuum. Each sulfonate group carries a negative charge, determining the overall charge of the complex. For the di-monomer case, E_tot(model) consists of the total energy of two isolated monomer molecules each with a single negative charge.

Additionally, we can use the above relations to estimate the ion adsorption energies with respect to hydrated ions, E_ad,hydr(ion), instead of with respect to ions in a vacuum using the literature-tabulated hydration free energies of the ions, E_hyd(ion), versus their reference states in a vacuum. The adsorption energy with respect to the hydrated state of each ion is:

$$E_{{\mathrm{ad,hydr}}}\left( {{\mathrm{ion}}} \right) = E_{{\mathrm{ad}}}\left( {{\mathrm{ion}}} \right) - E_{{\mathrm{hyd}}}\left( {{\mathrm{ion}}} \right).$$

The link to the common vacuum reference environment allows us to connect energy differences from theory with free-energy differences from experiment and create an approximate link to the hydrated reference state of the ion without the need to calculate the free energy of a hydrated free ion. The latter would be very difficult but, owing to the availability of experimental reference data, unnecessary. The main approximation we still make is to consider adsorption sites in a vacuum instead of in their polymer environment. However, as indicated in the main text, it seems that Ca²⁺ and Pb²⁺ in analogous local bonding situations have very similar adsorption energies with reference to the vacuum state in the present work as well as in a much more extensive assessment of different adsorption situations of divalent cations in past work^17,18.

Lead adsorption and leakage test

All glassware used in the lead leakage tests was cleaned and rinsed thoroughly with DI water. For the adsorption tests of Pb²⁺ by the CERs in aqueous solution, 150 mg CERs was poured into 50 ml of 1,000 ppm lead solution with constant stirring with an agitation speed of 120 r.p.m. A series of samples were extracted from the reaction mixture at different time intervals ranging from 1 to 120 min at different temperatures of 20, 40 and 60 °C. The samples were analysed by an ICP-MS Nexion 300D instrument to determine the Pb²⁺ concentration. The amount of lead adsorbed per unit mass of CERs at time t, q_t, was determined by the expression q_t = V(C₀ –C_t)/m, wherein C₀ and C_t (mmol l⁻¹) represent the initial concentration and after time t (min), respectively, m is the adsorbent mass (0.150 g) and V corresponds to the volume of the solution (0.05 l). The adsorption kinetics were then analysed by pseudo-first-order and pseudo-second-order models. For the lead leakage test from damaged perovskite solar modules, perovskite mini-modules were first damaged by an ice ball (5.08 cm diameter). For the water-soaking test, perovskite mini-modules were immersed in 200 ml DI water, and the Pb²⁺ concentration was measured (tenfold dilution) every 30 min using an ICP-MS Nexion 300D instrument. Before the test, the calibration curve for analysis was drawn by measuring standard solutions prepared by mixing the lead standard solution with a 2% HNO₃/DI water solution. For the water-soaking test, each damaged mini-module was placed in a Petri dish (diameter 15 cm) and 200 ml DI water at different temperatures (2, 20, 60 and 85 °C) was slowly poured into the dish from the edge to completely immerse the whole module. The lead concentration in the contaminated water was analysed every 30 min. For the water-dripping test, the damaged perovskite mini-modules were mounted in a funnel at an angle of ~30°. After that, deionized water with a pH of 7 or 4.2 was continuously dripped onto to the device surfaces using a dropping funnel at a rate of 5 ml h^–1 for 1 h (the set-up can be found in Supplementary Fig. 10). The lead-contaminated water was collected in centrifuge tubes and then further analysed by ICP-MS.

Lead simulation on perovskite solar panel

The lead adsorption on the carbon solar panel was simulated as follows. The solar panel was 198 cm long and 99 cm wide. A heavy rainfall of 50 mm h⁻¹ was considered in this simulation. We regarded the lead adsorption by CERs on glass as a diffusion-controlled process, and the Lagergren equation³⁵ was employed to describe the process: C₀ − C_t = C₀(1 − e^−kt), where k is the adsorption constant and was calculated to be 0.0892 based on the flowing-water results shown in Fig. 1d, t is the dwell time of water on the surface of the solar panel, C₀ is the initial lead concentration (14.3 ppb) and C_t is the lead concentration at dwell time t. The flow of rainwater over the solar panels was treated as a linear motion with an acceleration constant a of 1.875 cm s⁻² and its initial velocity v₀ was 0 cm s⁻¹. The simulation was performed with a home-made Python program. We divided the solar panel into 1980 × 990 pixels with each pixel having a size of 0.1 cm × 0.1 cm (denoted as p × p). We let (i, j) denote the pixel in row i and column j, with pixel (1, 1) situated in the top left corner of the panel. Then we calculated the water flow velocity from the top to the bottom of the panel. v_i is the velocity at the bottom of the ith row, where v_i = (v_i–1² + 2ap)^1/2. The time required for the water to flow from the top of the panel to the ith row was t_i, where t_i = [−v₀ + (v₀ + 2ap)^1/2]/a. Then we could calculate the concentration of leaked lead, C_i, flowing from the top of the panel to the ith row using $C_i = C_0{\mathrm{e}}^{-kt_i}$. The leaked lead concentration from the ith row and jth column pixel, C_i,j, should be related to all the pixels from the first to the ith row of the jth column. We also assumed the flow of rainwater downward from any point was the same at any time, and equal to the flow of rain falling upon each pixel, therefore $C_{i,j} = \mathop {\sum}\nolimits_{i = 1}^i {C_i/i}$. C_1,980,j is the lead concentration of all the pixels on the bottom row of the solar cell panel, which is the concentration of leaked lead in rainwater when the solar panel is completely damaged and subjected to rain. To simulate the water flow thickness, we calculated the average water flow velocity from the panel top to the panel bottom. v_i^a is the average velocity at the ith row, where v_i^a = (v_i–1 + v_i)/2. Then we could use the flow of rain falling upon each pixel to calculate the thickness of water flow on the ith row when the water flows from the top to the bottom of the solar panel. The water thickness on the ith row and the jth column pixel, T_i,j, should also be related to all the pixels from the first to the ith row of the jth column. Under the assumption that the flow of rainwater downward from any point was the same at any time, we obtain T_i,j = (i × 0.5 × 99 × p)/(99 × 3,600 × v_i^a).

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.