Hydrogen Bond-Enabled Carbon Dot-Based Network for Efficient Inverted Organic Solar Cells with Prolonged Photoelectrical-Mechanical-Moist Stability

Abstract

Achieving long operational stability is critical for advancing high-performance organic solar cells (OSCs) towards commercialization. Designing a charge transport layer with optimal electrical-physical-chemical compatibility is a pivotal strategy. Herein, we demonstrate an electron transport layer (ETL) by incorporating in-situ fluorinated carbon dots (FCDs) in a coherent polymer to construct the crystallinity-enhanced FCD network (^netFCD) via strong hydrogen bonding. Uniquely, ^netFCD-based ETLs offer very strong electrical and mechanical contact with their all-sided (cathode and active layer) interfaces. The reduced basicity feature of ^netFCD prevents unexpected interfacial reactions with non-fullerene acceptors. Meanwhile, ^netFCD-based ETLs enable barrier-free devices with high electron mobility and hole blocking. Using the strategically designed ^netFCD with the interesting features as “all-in-one” ETL, the inverted OSCs (i-OSCs) achieve 20.01% PCE (Certified PCE: 19.78%). The OSCs maintained 85% of their initial PCE after 1800 hours of operation under 80 ± 5% humidity while the flexible i-OSCs realized 95% of the initial PCEs after 10,000 bending cycles, advancing the operational (i.e., photoelectrical), mechanical, and environmental stability in OSCs with different structures. This work contributes to developing not only feasible bulk-to-interface interlayer designs but also green and low-cost materials to promote OSCs as sustainable and green energy sources.

Long-lasting organic solar cells require transport layers stable in electrical, mechanical and chemical aspects. Cui et al. fabricated a fluorinated carbon dot/polymer electron transport layer, which delivers high efficiency together with excellent humidity and bending stability.

Introduction

The inverted (i.e., n-i-p) organic solar cells (i-OSCs), with a structure of transparent cathode/electron transport layer (ETL)/active layer/hole transport layer (HTL)/ anode, exhibit favorable commercial relevance and improved operation stability in air compared to their conventional (p-i-n) counterparts^1–4. However, despite these structural advantages, the stability of i-OSCs still lags behind practical application requirements^5,6. To expedite commercialization, developing effective strategies for achieving i-OSCs with high efficiency, prolonged stability, and a comprehensive understanding of degradation mechanisms is crucial. Importantly, the degradation is widely recognized to originate from limited interface engineering when designing the ETL (i.e., bulk-to-interface interlayer design)^7,8. It causes the deterioration of ETL under different operation conditions, such as environmental factors (oxygen/moisture), operational conditions (prolonged light exposure/temperature), parasitic reactions, and mechanical stress^7–11.

The challenges of the simultaneous interface engineering in designing ETLs for improving device stability include: i) the external-environment-induced reactions. It commonly happens in different commonly used ETLs. For instance, Zinc oxide (ZnO), as a photocatalyst, exhibits interfacial defect reactivity under prolonged UV exposure and is prone to hydrolysis at its interface in high-humidity conditions^12,13. Organic small molecular ETL materials (e.g., perylene diimide (PDI)-based ETLs such as PDINN and PDINO) also suffer from poor thermal stability and photostability^14–16. ii) The intrinsic reactive nature. The detrimental interfacial chemical reactions between an interlayer and its adjacent layers require critical consideration. Specifically, strongly alkaline materials (e.g., PEIE) that exhibit a low degree of protonation can corrode sensitive active layers^17,18. The lone electron pair in polar groups (e.g., amino group) of ETL materials also interacts with the C = C linker in non-fullerene acceptors, forming irreversible reactions^19–22. iii) Mechanical stress-induced degradation. Inadequate interfacial adhesion typically results in delamination under the mechanical-bias-induced interlaminar stress, severely impeding the enhancement of device stability^9,23. Consequently, the bulk-to-interface ETL design is highly desirable to achieve the chemical stability (e.g., surface defect passivation and pH regulation), physical robustness (strong mechanical interactions with its adjacent layers), and electrical properties (e.g., barrier-free interface, high electron mobility, and hole blocking), implying the electrical-physical-chemical compatibility to develop all-in-one ETLs for efficient i-OSCs with prolonged operation stability.

Herein, we demonstrate an all-in-one ETL by strategically introducing a perfluoro sulfonic acid-based conductive polymer (Nafion®117) to construct in situ fluorinated carbon dots (FCDs) within a coherent polymer network (^netFCDs) with enhanced crystallinity via strong hydrogen bonding. This approach significantly enhances the interfacial adhesion among the bulk ^netFCD materials and between ^netFCD-based ETL and its adjacent layers. Through comparative studies, we reveal that hydrogen-bond interactions associated with semi-ionic C-F bonds and amino groups likely enhance electron mobility in the bulk ^netFCD film while simultaneously strengthening the electrical and mechanical contacts at all-sided interfaces. Meanwhile, the ^netFCD-based ETLs achieve a low lowest unoccupied molecular orbital energy level (E_LUMO) of ≈−4.00 eV for barrier-free electron extraction devices. The ^netFCD-based ETL also regulates the vertical phase donor/acceptor phase separation within the active layer through its heterogeneous surface energy distribution, thereby enhancing device performance. Notably, the reduced alkalinity (higher protonation) of ^netFCD mitigates irreversible chemical degradation at the ^netFCD/non-fullerene acceptor interface, significantly improving the operational stability of i-OSCs. With the dedicated design of the crystallization-enhanced ^netFCD with the interesting features for ETL, the i-OSCs exhibit a remarkable power conversion efficiency (PCE) of 20.01%. The T₈₀ lifetime of rigid i-OSCs exceeds 11,000 h at the maximal power point (MPP) with a humidity of 80 ± 5%. The ^netFCD-based flexible i-OSCs reached 18% PCE and retained 95% of their initial PCE after 10,000 bending cycles at a radius of 5 mm. The ^netFCD-based device performs better in terms of operational and mechanical stability than all inverted and normal structures of OSCs^24–32. Consequently, this work demonstrates not only robust, green, and low-cost hydrogen-bond-enhanced ^netFCD as efficient all-in-one ETLs, but also bulk-to-interface interlayer design for advancing the sustainable development of OSCs.

Results

Design and synthesis of FCD composite network

To achieve efficient ETLs for prolonging the OSC durability, Fig. 1a depicts the synthesis strategy of ^netFCD from poly(ethylenimine)-ethylene glycol (PEIE), Erythrosin B sodium salt (ErB), and perfluoro sulfonic acid-based conductive polymers (Nafion^®117, fluorine source) with facile one-pot hydrothermal reactions (Details are provided in the Methods section). Different from conventional applications of Nafion^®117 in fuel cells and water electrolysis for hydrogen production (primarily leveraging its high proton conductivity), the coherent polymer Nafion^®117 serves as both a fluorination source to form FCDs with semi-ionic C-F bonds on their surfaces and as the matrix to incorporate FCDs to construct a FCD network (^netFCD) as the ETL. It also mitigates the unexpected and irreversible chemical attack on non-fullerene acceptors by CD neutralization, ensuring the electrical-chemical-mechanical compatibility. Supplementary Fig. 1a shows the absorption curves of the ^netFCD and CD films. To better understand ^netFCD, the effects of dot dimension and chemical composition were studied. Figure 1b, c shows the schematic of the CD and ^netFCD, with the corresponding high-resolution transmission electron microscopy (HR-TEM) images (Fig. 1d, e) showing that the CDs and FCD share a similar average diameter of ~5 nm. Interestingly, FCDs exhibit larger lattice spacing (0.25 nm vs. 0.24 nm in CDs) and surface group-driven aggregation, which was determined by the FFT and line profile analysis as shown in Supplementary Fig. 1. Equally importantly, the self-assembled network structure in ^netFCDs is formed by incompletely cleaved fluorocarbon chains within the polymer matrix. Notably, the main fluorocarbon chain of Nafion^®117 is stable at high temperatures over 500 °C³³, while our process achieves fluorine ionization for CD functionalization via side-chain cleavage at 180 °C without breaking the Nafion^®117-based 3D network. The C 1 s signal of XPS (Supplementary Fig. 4) indicates that ^netFCD remains the main fluorocarbon chain. The Fourier transform infrared (FTIR) in Supplementary Fig. 2c, d shows that the new stretching vibration of covalent C-F bonds (1280 cm⁻¹) and semi-ionic C-F bonds (1120 cm⁻¹)³⁴ exist in ^netFCD compared with CDs. X-ray photoelectron spectroscopy (XPS; Fig. 1f–h and Supplementary Fig. 4) was used to comprehensively analyze multiple characteristic signals. They indicate the presence of semi-ionic C-F bonds at 684.4 eV in the F 1 s signal. This feature represents the characteristic bond of ^netFCDs, whereas pristine Nafion^®117 only shows a covalent C-F bond at ~688.5 eV. All the results indicate that Nafion facilitates the synthesis of FCD with abundant functional groups, and they are embedded in the coherent polymer matrix. The synchrotron-based grazing-incidence wide-angle X-ray scattering (GIWAXS) measurement (Fig. 1i–k) to characterize the films. In GIWAXS analysis, brighter polycrystalline diffraction rings indicate stronger crystallinity, owing to more efficient X-ray scattering from well-ordered crystalline regions. Accordingly^net, FCDs show enhanced crystallinity compared with pure CDs, despite the subtle visual difference. The strengthened crystallinity can be explained by the enhanced interaction between FCD and the polymer network, which can facilitate the charge transport in the film. The details will be discussed in the following paragraph.

Fig. 1. Material synthesis and characterization.

a Scheme of the synthetic process of CDs and ^netFCDs. b, c Schematic of the CDs and ^netFCDs. TEM (with HR-TEM image in the upper inset and red color to show the hydrogen bonds in the lower inset) images of CDs (d = 0.24 nm) (d),^net FCDs (d = 0.25 nm) (e). f–h XPS results of the F 1 s signal and full spectrum of ^netFCDs and Nafion^®117. 2D GIWAXS diffraction patterns of the thin films of CD (i),^net FCD (j) and Nafion^®117 (k) coated on ITO substrates. All GIWAXS plots use the identical color bar.

All-sided intermolecular interaction with adjacent layers

Firstly, all-sided interfacial contacts of ETL were investigated. We explored hydrogen bonds to establish the interactions of ETL with adjacent layers, including the underlying electrode (ITO) and top organic active layer. The schematic of the ^netFCD layer interacting with adjacent layers is shown in Fig. 2b. Figure 2a is the FTIR spectra of ITO^net,FCDs-dried, and ITO/^netFCD-dried (2 nm)^net.FCDs-dried and ITO/^netFCD-dried show a peak at 3300 cm⁻¹ and 3630 cm⁻¹, respectively, while there is no clear peak in the ITO control sample, indicating a stronger hydrogen bonding interaction of the ITO/^netFCD-dried surface. All samples were subjected to dry treatment by heating at 100 °C for 10 min. Combined with the FTIR and high binding energy shift of In 3 d peaks in Supplementary Fig. 5, we further verify that the hydrogen bonding interaction arises from the ITO/^netFCD-dried interface rather than water-induced interference. To elucidate the interaction between ETL and L8-BO, we conducted the ¹H NMR tests with CDs or ^netFCD (5 μL) added into the L8-BO deuterated chloroform (CDCl₃) solution. Figure 2c is the broadening and down shift of ¹H NMR singlet (δ) from δ1.55 (S0: L8-BO) to δ1.58 (S1: L8-BO+CDs) and then to δ1.64 (S2: L8-BO+^netFCD), which can be assigned to the hydrogen in the secondary amine to trail the hydrogen bond. Supplementary Fig. 4a, b shows that the peak shifts of L8-BO mixed with different concentrations of ^netFCD (from δ1.55 to δ1.78, Δ = δ0.23) are more remarkable than those of L8-BO mixed with CDs (from δ1.55 to δ1.64, Δ = δ0.09). The proton assignment for the δ = 1.55 ppm resonance of L8-BO is shown in Supplementary Fig. 6. The L8-BO interaction with ^netFCD is very likely due to the formation of intermolecular hydrogen bonds¹⁵. The same results were also confirmed by the variable-temperature ¹H NMR test (Supplementary Fig. 7c, d). Based on the above results, we propose a specific hydrogen-bond mechanism for S2. The semi-ionic C-F bond attracts a proton from –NH₂ in ^netFCD to form an F-H bond. This enables the F-H bond of ^netFCD to form a hydrogen bond with F or O in L8-BO (Supplementary Fig. 7e, f), resulting in a stronger interaction. Those results evidence that ^netFCD can achieve stronger F-H…F bonds due to the formation of a semi-ionic structure. This is particularly important in organic molecules as, in typical situations, the F atom in a stable covalent C-F bond prevents forming a hydrogen bond with SMA of active layers due to the lack of F-H interaction. Overall, the strong intermolecular hydrogen bonding with both the cathode and the active layer material has been confirmed, which can help strengthen the mechanical (adhesion) and electrical contacts.

Fig. 2. Interaction between ^netFCD and its adjacent layers.

a–c Model of interactions between the CD layer and adjacent layers through hydrogen bonding interaction. (a FTIR spectra of ITO^net,FCDs-dried, and ITO/^netFCD-dried; c.¹H NMR of CDs and ^netFCDs added in L8-BO solution.) d Surface energies at the millisecond and surface adhesion at the nanoscale of CDs and ^netFCDs obtained from the CA and PFQNM test. Surface adhesion mapping and line-cut profile curves extracted from PFQNM adhesion mappings along the vertical direction (Scan size, 500 ⅹ 500 nm), CDs (Rq = 0.66 nN) (e), and ^netFCDs (Rq = 1.14 nN) (f). g Absorbance spectra of 1 umol L8-BO in chloroform (CF) solution mixed with 100 ul methanol-dispersed interfacial materials (2 wt% PEIE, 20 mg/mL CDs, 20 mg/mL ^netFCDs), with the total volume of each solution being 2 mL after remaining 60 min. h, i N 1 s XPS results of CDs and ^netFCDs.

To deeply study the mechanical (adhesion) effect of ^netFCD on film, we examined the surface energy (γ_s) of CD and ^netFCD films through contact angle measurement (Supplementary Fig. 8) based on the Owens-Wendt model¹⁵. Apparently, the γ_s value (Fig. 2d and Supplementary Table 1) of ^netFCD (67.36 mN/m) is higher than that of CDs (63.44 mN/m), and thus better wetting properties for other materials on ^netFCD film. We also conducted AFM-based peak force quantitative nanomechanical mappings (PFQNM) to intuitively characterize surficial morphology and mechanical capacities of films in the nanoscale, as utilized by others^10,35. The surface adhesion force of CDs and ^netFCD is 25.08 nN and 33.18 nN, respectively, determined from the 2D adhesion mapping in Fig. 2e, f. The bright pink color indicates a strong adhesion force in parts of the area, and the ^netFCD film shows a large area of strong interaction. The standard deviation of surface adhesion (R_q) of ^netFCD film (1.14 nN) is much larger than that of the control CD film (0.66 nN), and the line-cut profile curves of ^netFCD exhibit more “hill-valley” features, implying that ^netFCD films exhibit a more heterogeneous surface energy distribution (He-SED). Besides, Supplementary Table2 summarizes that the ratio of low-adhesion peak area to that of high-adhesion peak area is 1.56 for CDs and 0.92 for ^netFCD, indicating ^netFCD film has a more balanced distribution between low and high surface energy at the nanoscale due to the hydrogen bonds.

Compared to CD, the ^netFCD exhibits reduced alkalinity (Supplementary Fig. 10a), suggesting its potential to mitigate degradation of non-fullerene acceptors. Figure 2g shows the absorption peak changes of 1umol L8-BO in chloroform (CF) solution mixed with 100 ul various methanol-dispersed materials (2 wt% PEIE, 20 mg/mL CDs, 20 mg/mL ^netFCDs) after remaining 60 min. To investigate the improved chemical stability, N 1 s signals of ITO/CDs and ITO/^netFCDs samples are shown in Fig. 2h, i. The feature centered at 402.1 eV is related to the protonated nitrogen of the amine group, and the content of protonated nitrogen of ^netFCDs (10.30%) is higher than that in CDs (4.96%), which can be explained by –HSO₃ in Nafion^®117 that facilitates the protonation of Nitrogen. The protonation strategy is an effective way for stable non-fullerene acceptor (NFA) based OSCs²¹. Compared with PEIE and CDs, the absorption peak of L8-BO mixed with ^netFCDs shows a negligible reduction in methanol, suggesting the chemical reaction between ^netFCD and L8-BO was suppressed.

Morphology modulation of the BHJ active layer by ETLs with nanoscale He-SED

Regarding the interaction influences of ETL films on the bulk-heterojunction (BHJ) active layer, we first investigated the root-mean-square roughness (R_q) by AFM measurement. The R_q of BHJ on CD and ^netFCD ETLs are 1.08 nm and 1.20 nm, respectively, as shown in Supplementary Fig. 11. According to the enhanced He-SED of ^netFCD film, we confirm that the increase in γ_s of the ETL induces the growth of grains and the large top-surface roughness of the active layers. Regarding the morphological variation of the buried surface of BHJ films, a 10 nm-thick BHJ film was coated on the ETL layers to mimic this condition. Nitrogen (N 1 s) as the characteristic element of L8-BO was used to assess its amount, as the XPS results are summarized in Supplementary Fig. 12 and Supplementary Table 3. The atomic ratio of N in the BHJ films deposited on ^netFCD films (3.98%) is larger than that of the CD case (3.61%), illustrating that the L8-BO content in the bottom region of the BHJ blend film increases for the ^netFCD case with the higher γ_s. It can be understood that the higher γ_s of L8-BO relative to PM6 and the change of He-SED could affect the L8-BO distribution in the blends³⁶, as also evidenced by the TOF-SIMS results (Fig. S23). Supplementary Figure 13 shows the influence of different ETLs on the molecular stacking orientation in the BHJ films by GIWAXS characterization. Our results show that the ^netFCD film enhances the ordering of π-π packing of the BHJ layer, advancing the charge transportation in the devices. Besides, as more L8-BO acceptors will be distributed near the ^netFCD ETL surface, the electron extraction via ^netFCD ETL to the cathode is promoted.

In addition, we study the evolution of the peak absorption of the active layer under different tear-off times to understand the influence of ETL adhesion on the upper blend layer (100 nm). In the comparison of BHJ coated on ZnO or CDs, the good adhesion property of the BHJ layer deposited on ^netFCD layers, as evidenced by negligible absorption peaks change on the top surface (Supplementary Fig. 14), suggests that the BHJ layer is not likely to undergo delamination from the underlying layer. The inserted image of Supplementary Fig. 14d shows that the reduction of the absorption peak of PM6 is more than that of L8-BO on the top surface, confirming a change in the vertical phase separation in the BHJ films. This change is expected to reduce charge recombination at the D/A interface³⁷. Consequently, all of the results indicate that the BHJ layer on ^netFCDs ETL exhibits higher ordering in the π-π stacking and enlarged vertical phase separation with larger domain size (Fig. 3a), contributing to the improvement of device performances.

Fig. 3. Morphology and optoelectronic properties of OSCs.

a Schematic of the interaction between ^net FCD and L8-BO. b UPS spectra of ITO, ITO/CDs, and ITO/^netFCD. c Current-voltage (I-V) curves of devices with a structure of ITO/CDs or ^netFCD/Ag. d W_F of ITO electrodes modified by the ETLs and energy levels of CDs^net, FCD, Y6, L8-BO, PM6, and D18. e The current density (J)-voltage (V) curves of optimized PM6: D18: L8-BO based inverted OSCs under AM 1.5 G (100 mW/cm²) illumination. f EQE and integrated J_sc of the corresponding devices. g J-V curves of devices with different ETLs in the dark. h J-V curves of the optimized i-OSCs under AM 1.5 G (100 mW/cm²) illumination. i EQE curves and integrated J_sc of the corresponding devices. All of the optoelectronic results of OSCs are based on PM6: D18: BTP-eC9:L8-BO system.

Photovoltaic properties and performances

To understand the electrical properties of CD and ^netFCD films, an ultraviolet photoelectron spectroscopy (UPS) was conducted. The W_F of ITO, ITO/CDs, and ITO/^netFCDs are −4.70 eV, −3.94 eV, and −3.97 eV, respectively, as shown in Fig. 3b. The tendency indicates that ^netFCD regulates the W_F of the ITO electrode for better ohmic contact. Together with the optical bandgap experiment, we determined the E_HOMO and E_LUMO, as shown in Fig. 3c and Supplementary Fig . 15. Both the experimental and theoretical results indicate that ^netFCDs exhibit lower E_LUMO from −3.73 eV (CDs) to −4.00 eV (^netFCDs), confirming the effect of fluorination on the carbon core^38,39. The E_LUMO of ^netFCDs is similar to that of NIR-SMAs, improving band alignment at the cathode interlayer and establishing a barrier-free device for better electron extraction. In addition, the higher E_HOMO enables the ETL to effectively block the holes from donors (≈−5.50 eV)^13,40. The I-V curves for conductivities of CDs and ^netFCDs (Fig. 3d) are 8.86 × 10⁻⁴ S cm⁻¹ and 1.55 × 10⁻³ S cm⁻¹. The µ_e was also extracted by the space-charge-limited current (SCLC) method in electron-only devices (ITO/ZnO/CDs or ^netFCDs/PDINN/ Ag). From J-V curves (Supplementary Fig. 16), µ_e of CDs and ^netFCD are determined as 3.86 × 10⁻⁴cm²V⁻¹ s⁻¹ and 4.95 × 10⁻⁴cm²V⁻¹ s⁻¹, respectively. The photoluminescent (PL) spectra and transient PL (TRPL) (Supplementary Fig. 2) show that the decreased intensity of the ^netFCD peak and faster emission decay of ^netFCDs may contribute to the strong electronegativity of the surface F atom for photoexcited electron transfer and effective charge separation^41,42. Consequently^net,FCDs offer better µ_e and extraction as compared to the pure CDs case, which could be explained by the good crystallinity and increased charge concentration for superior electron transport of CDs.

The performance of inverted OSCs with a structure of ITO/ETLs/PM6:D18:L8-BO/MoO_3/Ag is shown in Supplementary Fig. 17. The transmittance of ITO/^netFCD is over 93% as shown in Supplementary Fig. 2b. Supplementary Fig. 18 shows the chemical structures of active layer materials. L8-BO acceptor has a low E_LUMO at −3.90 eV with a negligible energy offset to FCD ETL, causing a more positive electron transport path¹³. The current density-voltage (J-V) curves of OSCs are shown in Fig. 3e. The photovoltaic parameters are summarized in Table 1. The ^netFCD-based OSC (PM6:D18:L8-BO) delivers an impressive PCE of 19.17%, which is much better than that of CDs (PCE 18.04%), ZnO (PCE 17.14%), and SnO₂ (PCE 16.22%) based devices. The external quantum efficiency (EQE) curves and integrated currents are shown in Fig. 3f and Table 1, respectively, and agree well with the J_sc results (the deviation <3%). IQE (Supplementary Fig.19c) curves indicate that the J_sc improvement is non-optical in origin. The ^netFCD-based OSCs expressively depict higher EQE response in the 550-800 nm range. The J-V curves in the dark are shown in Fig. 3g. Overall, the results show that the ^netFCD-based OSC has lower series resistance, larger rectification, and better diode characteristics in comparison to ZnO- and CD-based OSCs, implying the superior charge transport property of the ^netFCD ETL.

Table 1.

Summarized photovoltaic parameters of OSCs based on ternary BHJ with different ETLs measured under AM 1.5 G (100 mW/cm²) illumination

BHJ	ETL	Voc (V)	Jsccal) (mA/cm–2)	FF (%)	PCEb) (%)
PM6:D18:L8-BO (Glass)	ZnO	0.874	26.11 (24.99)	75.35	17.14 (16.79 ± 0.19)
	SnO2	0.872	26.24 (25.18)	71.37	16.22 (15.98 ± 0.17)
	CDs	0.876	26.68 (26.26)	77.37	18.04 (17.67 ± 0.17)
	netFCDs	0.880	27.23 (26.67)	80.02	19.17 (18.77 ± 0.18)
PM6:D18:L8-BO (PEN)	CDs	0.873	25.87 (24.39)	76.06	17.18 (16.96 ± 0.20)
	netFCDs	0.880	26.26 (24.99)	77.90	18.00 (17.87 ± 0.17)
PM6:BTP-eC9:DY-IDT(Glass)	netFCDs	0.870	28.64 (28.17)	79.64	19.84 (19.44 ± 0.35)
PM6:D18:BTP-eC9:L8-BO (Glass)	netFCDs	0.890	28.15 (27.28)	80.05	20.01 (19.73 ± 0.22)

^cal)The integrated J_sc from EQE measurement. ^b) The average PCEs with deviation were calculated from 15 samples.

To further understand the excellent performance of ^netFCD-based OSCs, their ability to charge extraction and recombination was studied. The photocurrent density-effective voltage (J_ph-V_eff) characteristics³⁶ were measured, as shown in Supplementary Fig. 19b. The higher V_eff means all of the photo-induced excitons are dissociated, and carriers are collected. In our case, the saturation current (J_sat) was obtained when V_eff = 3.0 V, and the charge dissociation probability (P(E,T)) was also calculated from the equation of J_ph/J_sat. Under short circuit conditions, we determined the P(E,T) values of ZnO, CDs, and ^netFCDs ETLs as 98.7%, 98.7%, and 99.0%, respectively. Moreover, the ^netFCD-based device shows higher in low voltage, indicating that less driving force is needed to extract the carriers in ^netFCD-based OSCs. The charge recombination behavior of devices has also been studied by the dependence of J_sc and V_oc on light intensity (P_L). Supplementary Fig. 19c shows the J_sc-P_L curves with the formula J_sc∝P_L^s2. Since the index is close to 1, it indicates that no obvious bimolecular recombination exists in ^netFCD-based OSC. The trap-assisted recombination could be investigated from V_oc-P_L curves (Supplementary Fig. 19d). With the relationship of V_oc∝skT/q ln(P_L), the slope between kT/q and 2kT/q represents trap-assisted recombination¹⁹. The devices based on ZnO-, CDs-, and ^netFCDs- display slopes of 1.10 kT/q, 1.08 kT/q, and 1.01 kT/q, respectively, demonstrating that the trap-assisted recombination was effectively suppressed in the devices²⁰. To explore the universality of the ETLs, we further applied PM6:Y6 (binary) (Supplementary Table 4), PM6:BTP-eC9:DY-IDT (ternary), and PM6: D18: BTP-eC9:L8-BO (quaternary) systems to fabricate OSCs, the PCE of the quaternary system has reached 20% (certificated PCE: 19.78%, Supplementary Fig. 24) among the highest PCEs for OSCs to date, and the corresponding EQE spectrum is shown in Supplementary Fig. 20 and Fig. 3. The results show that ^netFCD ETL is compatible with low E_LUMO acceptors.

Device stability for long-term application

To support the sustainable development of OSCs, we evaluated the operational stability of different ETL-based devices under continuous 1-sun illumination (AM 1.5 G, 100 mW/cm²) at the maximal power point (MPP), highlighting their potential for long-term, eco-friendly application. The efficiency decay was monitored in encapsulated devices under an air condition of 80 ± 5% humidity. The average PCE as a function of time decay curves of varied ETL-based OSCs with error bars is shown in Fig. 4a and collected from three independent experiments for reliability (Supplementary Fig. 21). The ^netFCD-based OSC retains 85% of its initial PCE after 1800 h illumination and enables an extrapolated T₈₀ lifetime of >11,000 h⁴³, which is much better than that of control ZnO-based and CDs-based packaged devices. The results demonstrate that the ^netFCD-based i-OSCs have the supreme realistic condition operation stability (Fig. 4d and Supplementary Table 5). To better understand the performance, we study the chemical interaction of ETLs in different conditions. For waterproofness ability, Fig. 4b, c shows the evolution of PCE with ITO/ZnO/BHJ, ITO/CDs/BHJ, and ITO/^netFCDs/BHJ half-cells aged in water for many hours and then evaporated Ag to complete full devices. The ZnO-based devices exhibited 60% initial PCE loss when the half-cell was soaked in water for 3 h, due to the chemical dissolution of ZnO (the inserted optical images), with the equation¹² of $\mathrm{ZnO}+\mathrm{n}{\mathrm{H}}_2\mathrm{O}+2{\mathrm{h}}^{+}\to {\left[\mathrm{Zn}{\left(\mathrm{OH}\right)}_{\mathrm{n}}\right]}^{\left(2-\mathrm{n}\right)}+\frac{1}{2}{\mathrm{O}}_2+\mathrm{n}{\mathrm{H}}^{+}$. However, the devices using nanoparticles as ETL show better water-immersed stability, with the efficiency maintaining 71% (CDs-based) and 90% (^netFCD-based) of the initial PCEs. Overall, these results support that the enhanced adhesion between ETL/ITO and ETL/BHJ interfaces can improve device stability, which makes ^netFCD-based devices capable of on-land and underwater applications^12,44.

Fig. 4. Device stability.

a Normalized PCE plotted against aging time under 1-sun illumination from a Xenon lamp at MPP tracing under air condition (ISOS-L3). Data from 3 independent devices; error bars denote standard deviation (SD). b Illustration for the water-immersion stability test of the studied OSCs (upper shows immersing the ITO/ETL/BHJ structure in water, then adding MoO₃ and silver to fabricate the device). c The device stability using different ETLs and optical microscope images of the BHJ film coated on the different ETLs aged in water. Data from 10 independent devices; error bars show SD. d PCE versus operation lifetime of previous works and this work. e Normalized PCE of CD- and ^netFCD-based flexible solar cell as a function of bending radius is 5 mm (F is Force). Data from 10 independent devices; error bars show SD. f Devices PCE versus bending cycles after 1000 times of previous works, and this work is based on flexible ITO electrodes.

Benefiting from the advantages of high adhesion and the room-temperature process to form ETL, we fabricate the flexible devices with the structure of PEN/ITO/CDs or ^netFCDs/BHJ/MoO₃/Ag. Compared to CD-based devices, the ^netFCD-based OSCs (Supplementary Fig. 22 and Table 1) achieved a higher PCE of 18% with a V_oc (from 0.87 to 0.876 V), a J_sc (from 25.6 to 25.8 mA/cm²), and an FF (from 74.2 to 77.5%), which is one of the highest flexible OSCs reported in the articles. Figure 4e, f shows that the CDs- and ^netFCD-based devices remain at 78% and 90% of the initial efficiency after 2,000 bending cycles at a radius of 5 mm. Moreover, the ^netFCD-based device maintained >95% of the original PCE after 10,000 bending cycles with the consideration of neutral plane effects^24–32, implying that the enhancement of adhesion results in good mechanical stability of ^netFCD-based OSCs. Overall, the ^netFCD-based i-OSCs achieve better operational and mechanical stability than all inverted and normal structures of OSCs^24–32.

Discussion

We demonstrate a strategy to establish hydrogen-bond-driven FCDs embedded into the coherent polymer network to form the ^netFCD structure, serving as the “all-in-one” ETL of i-OSCs. The reinforced interfacial interactions of ^netFCD-based ETL to the adjacent active layer and electrode improve the charge collection and extraction abilities, respectively. The enhanced crystallinity of ^netFCD offers a high electron mobility of 4.95 × 10⁻⁴cm²V⁻¹ s⁻¹. The ^netFCD also possesses a lowered E_LUMO of ≈−3.91 eV and modulates the W_F of ITO from −4.70 to −3.88 eV for favorable band alignment. In PM6:D18:L8-BO inverted OSCs with ^netFCD-based ETL, the matched E_LUMO of the ETL with the L8-BO acceptor exhibits a barrier-free structure, improving the charge extraction of devices. With the excellent interfacial compatibility and good device energy-level alignment, the ^netFCD-based i-OSCs achieve an optimal PCE of 20.01% in rigid and 18% in flexible i-OSCs. Additionally, the ^netFCD-based devices illustrate excellent stability in various aspects, including chemical stability, waterproofness, and mechanical flexibility, rendering them suitable for on-land and dive-into-water applications. The work demonstrates not only robust and stable crystallization-enhanced ^netFCD as efficient, non-toxic all-in-one ETLs but also contributes to establishing bulk-to-interface interlayer engineering approaches to promote OSCs as sustainable green energy resources.

Methods

Materials

All reagents and materials, unless otherwise specified, are purchased from Sigma-Aldrich or Alfa Aesar and used without further purification. PM6, D18, Y6, and L8-BO are purchased from Solarmer Materials Inc. DY-IDT is supplied by Prof. Jen’s group.

Material synthesis

For CDs, the synthesis method is 2 mL PEIE and 26 mg ErB were dissolved in 8 mL DI water, then the solution was transferred to a 20 mL Teflon-lined autoclave under 180 °C heating for 4 hours. For ^netFCDs, 2 mL PEIE, 26 mg ErB, and 200 μL Nafion^®117 were dissolved in 8 mL DI water. Then, the solution was transferred to a 20 mL Teflon-lined autoclave under 180 °C heating for 6.5 hours. The products were firstly purified by high-speed centrifugation (8000 x g, 5 min) to remove aggregates and large particle impurities, then the supernatant was collected and dialyzed in a 1000 Da dialysis membrane for 24 h. Finally, they are freeze-dried to a solid with dialysis and stored at 4 °C for further measurement and use. The one-pot hydrothermal reaction is convenient for further scale-up if needed. The yield of the product (^netFCDs) is approximately 74%, and the estimated price of ^netFCDs is about 0.55RMB/(mg/mL).

Encapsulation procedure

A glass cover slide was placed over the device and sealed/pressed to the substrate with an encapsulant to provide edge sealing, then cured under 450 nm UV light for 5 min.

Device fabrication and measurements

Patterned ITO-glass substrates are used as the cathode of organic solar cells. The ITO-coated glass substrates are cleaned by sonication in detergent, deionized water, acetone, and isopropyl alcohol and dried in a nitrogen stream. The prepared CDs and ^netFCDs solutions with different concentrations were spin-coated on ITO substrates, then heated at 100 °C for 10 min to remove residual solvent. The substrates are then transferred into a nitrogen-filled glove box. The PM6:Y6 (1:1.2 w/w) 1-chloronaphthalene (1-CN) (97:3 v/v) solution is spin-coated on different cathode interlayers and followed by 10 min thermal annealing (100 °C). The PM6:D18:L8-BO (0.8:0.2:1.2 w/w, 14.3 mg/mL) 1,8-Diiodooctane (DIO) (95:5 v/v) solution is spin-coated on different cathode interlayers, followed by 85 °C thermal annealing for 10 min. The PM6:BTP-eC9:DY-IT (1:0.8:0.4 w/w, 14.3 mg/mL) 1,8- Diiodooctane (DIO) (97:3 v/v) solution is spin-coated on different cathode interlayers, followed by 90 °C thermal annealing for 5 min. The thicknesses of active layers are around 110 nm. An 8 nm MoO₃ layer and a 100 nm Ag layer are subsequently evaporated through a shadow mask to define the active area of the devices (8.00 mm²) and form the top anode. All the device fabrication processes are carried out in an N₂-filled glove box. The J-V characteristics are measured with a Keithley 2400 source meter under 1 sun, AM 1.5 G spectrum from a solar simulator (Enlitech; 100 mW cm⁻²). The solar simulator illumination intensity is determined using a monocrystal silicon reference cell calibrated by the National Renewable Energy Laboratory (NREL). EQE is recorded on a commercial EQE measurement system (Enlitech, QE-R3011, Taiwan).

The photostability of OSCs was evaluated under 1-sun illumination (100 mW cm⁻²) in air (50 ± 5 °C/RH 80%), following the ISOS-L-3 protocol. A Xenon lamp (ABET Technologies, USA) was used without UV filtering. The light intensity was calibrated using a certified silicon reference solar cell (K801SK302, McScience). Full encapsulated devices (ITO/ETL/active layer/MoO₃/Ag) were prepared by covering the device with a UV-curable adhesive and a cover glass, followed by UV curing (450 nm) for 10 min. Devices were tested under continuous maximum power point (MPP) tracking for 1800 h. Each data point represents the mean ± standard deviation of three independently prepared devices.

Reporting summary

Further information on experimental design is available in the Nature Portfolio Reporting Summary linked to this Article.

Author contributions

W.C.H.C. supervised the execution of the project. M.C. and W.C.H.C. conceived this project and designed the experiments. M.C. fabricated and characterized the network fluorinated carbon dots and photovoltaic cells. H.G. and A.K.-Y.J. provided DY-IDT NFA and offered suggestions about device optimization and characterizations. H.W carried out the GIWAXS measurement. M.C. and W.C.H.C. drafted the manuscript. All authors participated in the discussions for manuscript preparation.

Peer review

Peer review information

Nature Communications thanks Ergang Wang and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data that support the findings of this study are available from the corresponding author on request. The related data generated in this study are provided in the Supplementary Information/Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76944-z.