Suppressing electron-phonon coupling and energy loss in organic solar cells by modulating donor-acceptor penetrated-interface

Abstract

Organic solar cells (OSCs) achieve 21% efficiency, yet non-radiative energy loss (qΔV_nr) remains a critical barrier to further improve the open-circuit voltage (V_OC). This loss is primarily governed by the optoelectronic properties of interfacial CT states, yet the precise role of electron-phonon coupling (EPC) is not fully resolved. Through analysis of three all-polymer OSCs and four small molecule acceptor (SMA)-based OSCs, we identify two donor-acceptor (D-A) interfacial mixed phases that foster two distinct CT states, establishing efficient charge generation. These two phases emerge from amorphous D-A entanglement, termed as Entangled (E-) interface, and the penetration of acceptor quasi-aggregates into donor polymer matrix, termed as Penetrated (P-) interfaces. The P-interface exhibits inherently weaker EPC than that of E-interface since the suppressed intramolecular interaction. As the results, the P-interfaces, governing all-polymer OSCs, achieve a significant reduction of ~60 meV in qΔV_nr compared to E-interface dominated SMA-based OSCs. The incorporation of PA into SMA system as guest component modulates the population of P-interface reducing the EPC and then enhancing V_OC. Overall, our work suggests that modulating the population of P-interfaces to suppress EPC is a viable strategy for reducing non-radiative voltage loss and overcoming the efficiency bottleneck of organic solar cells.

The electron-phonon coupling in organic semiconductors is key to improving performance for devices. Here, the authors show that modulating the penetrated donor-acceptor interface with low electron-phonon coupling, reduces non-radiative voltage loss for high-performance organic solar cell devices.

Introduction

State-of-the-art NFA based systems, such as PM6-Y6, achieved high V_OC (>0.85 V) despite minimal energy offsets (driving force) between the energy of excited singlet state (E_S1) and charge transfer (CT) state (E_CT), denoted as ΔE_S1-CT. these systems maintain over 90% charge generation efficiency, challenging the two-state model that predicts a large ΔE_S1-CT is necessary to minimize qΔV_nr. To address this discrepancy, Nelson et al. and Gao et al. proposed a three-state model emphasizing the role of S₁-CT hybridization and thermal population. This model introduces the concept of intensity borrowing from the high-oscillator-strength E_S1, which enhanced the luminescence of the CT state and reduce the qΔV_nr, as^1–4:

$$\mathrm{\Delta }{V}_{\mathrm{nr}}=-\frac{k_{\mathrm{B}}T}{q}\ln \left({\mathrm{EQE}}_{\mathrm{EL}}\right)=-\frac{k_{\mathrm{B}}T}{q}\ln \left(\beta \frac{k_{\mathrm{r}}}{k_{\mathrm{r}}+k_{\mathrm{nr}}}\right)$$ 1

Here, the external electroluminescence quantum efficiency (EQE_EL) of devices is determined by the radiative (k_r) and non-radiative (k_nr) recombination rate constants, and the out-coupling factor (β) of the device, assumed to be 0.2^1,5. One guideline for reducing $\mathrm{\Delta }{V}_{\mathrm{nr}}$ involves enhancing the S₁ and CT states hybridization for improving CT state emission by intensity borrow mechanism. Therefore, the nature of CT states should be discussed carefully⁶. Additionally, the coexistence of multiple interfacial CT states in recent BHJ systems has introduced diverse charge generation pathways⁷. Salleo et al. and Amassian et al. directly link NFA blend morphology and structural order/disorder to multi-CT states distributions and associated non-radiative recombination process via systematic compositional tuning^8–11. Despite these advances, the mechanistic understanding of how interfacial CT states govern charge generation dynamics remains incomplete, particularly in emerging polymer donor-polymer acceptor (PD-PA) systems. Moreover, definitive morphological signatures identifying specific interfacial architectures and corresponding CT states are lacking.

Aside from those electronic properties, electron-phonon (vibrational) coupling (EPC) of CT states has received limited attention. EPC characterizes the interplay between molecular geometry deformations and electron transfer processes¹². Specifically, it describes the degree of excitons, CT excitons (CTEs), and polarons undergo phonon scattering, subsequently dissipating absorbed energy as heat from lattice vibration through non-radiative decay pathways prior to charge separation and collection^13–15. The strong EPC leads to rapid non-radiative decay, a well-documented limitation in fullerene-based OSCs^16,17. Minimizing EPC is thus a promising strategy to suppress non-radiative recombination and reduce voltage losses^18,19. Approaches such as enhancing orbital delocalization¹⁶, which is more common in polymer acceptor than small molecular acceptors, or employing molecular design strategies, including the introduction of noncovalent interaction between NFAs non-rigid end group²⁰, asymmetric side chain in NFAs promoting the more planar skeleton²¹, and solid additive to planarize the lattice vibration¹⁵, have been shown to lower EPC. Jen et al. utilized Huang–Rhys factor (S) to quantify the EPC of Y6 as 1.07 and introduced the diluted strategy to reduce the EPC and ΔV_nr^22,23. Nevertheless, most EPC studies have focused on pure phases and materials, leaving the influence of EPC at complicated D-A interfaces, where exciton dissociation and charge generation occur, largely unexplored.

In this work, we investigat two emerging PD-NFA systems, PD-SMA system and PD-PA system, with contrasting aggregation behaviors that govern their bulk heterojunction (BHJ) optical and morphological landscapes. By analyzing four PD-SMA system and three PD-PA system, the PD-SMA systems tend to form high-purity acceptor domains and disordered mixed phases, whereas the PD-PA systems exhibit a reversed morphology comprising disordered pure domains but ordered intermixed phases, consistent with previous findings by Chow et al.²⁴. We uncover distinct phase evolution pathways under compositional variation^11,25. SMA systems undergo rapid phase separation, while the PA systems exhibit gradual transitions from the fully amorphous phase (Entangled (E)-interface), to the quasi-aggregates and penetrated polymer chain filled phase (Penetrated (P)-interface), and ultimately to the crystalline pure phase. The P-interface exhibit slight edge-on molecular orientation relative to the pure acceptor domains. Critically, time-dependent density functional theory (TD-DFT) calculations and experimental results confirm that such P-interfaces exhibit intrinsically weaker EPC compared to disordered E-interfaces²⁶. Besides, P-interfaces also facilitates efficient charge generation, as seen from the kinetics of P-interface dominated PA systems. In performance, simulated energy loss analysis reveals that these low-EPC P-interfaces reduce non-radiative energy losses by ~58 meV, aligning with experimental reduction of energy loss (~55 meV) of PA system. The achieved V_OC increase in PA-SMA ternary system (D18-PY-IT-eC9) supports the strategy to modulate the population of P-interface by incorporation PA into SMA system. This work not only identifies the characteristics of P-interfaces but also elucidates the P-interfaces as multifunctional interfacial architectures that both suppress voltage losses and enhance charge generation. These findings offer valuable design principles, specifically highlighting the importance of optimizing P-interface formation to enhance the performance of organic solar cells (OSCs).

Results

Basic optical properties and energy level alignment

Figure 1a shows the chemical structures of polymer donors (PDs) D18 and PM6, small molecular acceptors (SMAs) eC9 and Y6, and polymer acceptors (PAs) PY-IT and PY-DT. All of the D-A combinations have been widely used in high-efficiency SMA-based and all-polymer organic solar cells (OSCs)²⁷. The energy levels for the representators, D18 (PD), eC9 (SMA), and PY-IT (PA) are determined by ultraviolet Photoemission Spectroscopy (UPS) and the intersection of the normalized absorption and photoluminescence (PL) spectra (Fig. 1b, c, Supplementary Figs. S1, S2)²⁸. The HOMO level and E_g are nearly identical for PA (−5.45 eV, ~1.5 eV) and SMA (−5.45 eV, ~1.48 eV), differing by only ~20 meV in E_g (Supplementary Figs. S1, S2). Deconvolution of neat PL spectra reveals a more pronounced disordered phase in PA than SMA, consistent with PA’s stronger intermixing tendency (Supplementary Fig. S3)⁶. To examine optical property evolution, blend films are prepared with the compositional variations of D-A ratios, 10-1, 10-2, 10-5, 10-10 and 10-15. Increasing acceptor concentration induces a redshift in both absorption and PL spectra. This shift is significantly larger in SMA-based blends (~51 meV from 10-1 to 10-15) than in PA-based blends (~11 meV), reflecting stronger electronic interactions in SMA systems (Supplementary Figs. S4, S5, Supplementary Table S1)²⁹. These spectral shifts may arise from order-to-disorder phase transitions, which are more prominent in SMA-based blends. At low acceptor concentrations, films comprise primarily disordered domains with intermixed D-A interfaces¹¹. Higher acceptor concentrations promote ordered pure domains, causing the observed redshifts (Supplementary Fig. S6). The pronounced shifts in SMA blends suggest enhanced intermolecular coupling from intrinsic SMA self-aggregation, amplifying contrasts between disordered and ordered phases³⁰. In contrast, minimal PA shifts imply that intra-chain wavefunction delocalization yields intrinsically smaller E_g²⁴. We perform the in-situ spectroelectrochemistry (SEC) measurement, the in-situ cyclic voltammetry with UV-vis-NIR spectroscopy, to determine the energy level of interface under different D-A ratio^31–33. By comparison, the disordered phase (interface) exhibits shallower HOMO depth than the pure acceptor (represented by pristine acceptor) and smaller ΔHOMO difference (ΔHOMO) between donor and acceptor, indicating a uphill energy landscape transition from pure domain to interface, also supporting by the deconvolution of UPS spectra (Supplementary Figs. S7–S11)^6,34,35. Interestingly, the PA system shows continuously shallowed ΔHOMO, which might link to the morphology evolution³². Notably, increasing acceptor ratios amplifies energetic order in both in PA and SMA system as the growth of ordered pure domain. The larger Urbach energy (E_urbach) in SMA system at low D-A ratio indicates more disordered phase than PA system (Supplementary Fig. S12)³⁶. However, mapping the energy landscapes of these systems remain challenging due to unclarified electronic properties of D-A interfaces, the critical sites for exciton dissociation and charge generation. Given the substantial impact of interfacial properties on device performance, systematic investigation of these interfaces is essential for further optimization.

Fig. 1. Chemical structures of photoactive materials and their physical properties.

Chemical structures of a PY-IT and PY-DT, b D18 and PM6, c eC9 and Y6. d Energy levels of PY-IT, PY-DT, D18, PM6, eC9 and Y6, e Absorption of PY-IT, PY-DT, D18, PM6, eC9 and Y6.

The morphology evolution of entangled- and penetrated-interface in blends

To gain deeper insight into the interfacial morphology of the blends, we characterized miscibility and phase behavior. As shown in Supplementary Table S2 and Supplementary Fig. S13, the PD-PA blend exhibits a smaller Flory-Huggins interaction parameter (χ) than PD-SMA. This indicates stronger thermodynamic miscibility and enhanced donor–acceptor percolation at the interface in PA-based blends. We employ grazing-incidence wide and small-angle X-ray scattering (GIWAXS and GISAXS) to analyze molecular packing and domain size, focusing on interfacial morphology (Supplementary Figs. S14–S19)³⁷. As the acceptor ratio increased, SMA blends develop tighter molecular packing (smaller d-space) and the longer crystal coherent length (CCL) in both lamellar and π-π stacking (Supplementary Fig. S20a). PA blends exhibit extended CCL in two types of stacking but looser π-π stacking due to PA’s inherently less compact structure (Supplementary Fig. S20b). These trends suggest that SMA favors crystalline, tightly packed pure domains, whereas strong interfacial percolation in PA promotes larger intermixed domains and looser pure domains. This is further corroborated by the Flory-Huggins two-phase diagram, which shows a higher D-A ratio threshold for the 1-phase to 2-phase transition in PA systems (Supplementary Fig. S21)^24,38.

We now turn to stacking orientation preference by extracting scattering intensity distribution, revealing the increased face-on orientation with higher acceptor ratio. The edge-on/phase-on ratio (A_z/A_xy) for lamellar stacking increases (Supplementary Fig. S22), and the peak width (FWHM_π-π) of face-on π-π stacking narrows (Fig. 2a, b)³⁹. Additionally, the evolution of Herman’s Orientation parameter (H) confirms the same trends, transition from edge-on to face-on orientation, supporting by the HUMO (IE) level shifting measured from in-situ SEC (Fig. 2a, b, Supplementary Figs. S9, S22, and Supplementary Note S1)⁴⁰. This phase transition can also be observed from charge transport mobility evolution both in In-plane (IP) and out-of-plane (OOP) direction measured from OFET and OPV devices (Supplementary Fig. S24). The detailed comparisons of FWHM_π-π, A_z/A_xy ratio and H parameter reveal that SMA system transit from edge-on to face-on packing rapidly and significantly at low D-A ratio (H_10-1 = −0.23 to H_10-5 = 0.7), stabilizing later (H_10-15 = 0.79). In contrast, the PA system transitions gradually. Notably, at low ratio of 10-1, the broader FWHM_π-π in SMA system may donate the more disordered nature of intermix region in SMA system. We quantified intermixed (isotropic) regions by the intermix factor (Λ), which, in brief, described the ratio of intermix scattering intensity to total scattering intensity (Supplementary Note S1). The larger Λ in PA system (Λ_10-15 = 0.18, Fig. 2a) signifies retained intermixed phase, while lower Λ in SMA reflects greater crystallization (Λ_10-15 = 0.07, Fig. 2b).

Fig. 2. Morphology analysis for the E- and P-interface.

a, b The full width (FWHM) at half maximum (histograms), Herman’s orientation parameters (H), and intermix factor (Λ) of π-π stacking peak at OOP direction of PA system and SMA system. c The lamellar features in blend films are a combination of the individual lamellar features for donor and acceptors, normalized to neat donor films. The higher rDoC for actual lamellar feature compared to expected rDoC indicates that the mixed phase contains the packed lamellar order in excess of expected. d The edge-on diffraction intensity (of diffraction ring) ratio to lamellar stacking peak for two systems.

At low acceptor ratios (10-1 and 10-2), both PA and SMA systems initially form entangled amorphous regions but minimal pure acceptor regions. Subsequently, pure regions in SMA system grows rapidly until reaching a threshold ratio of 10-5 and then stabilizes, whereas the PA system undergoes a steady gradual transition. This stepwise evolution underscores the transitions from intermixed region to pure domains in both systems. The difference between the relative degree of crystalline (rDoC) of actual and expected lamellar (ΔrDoC) of blends demonstrates the presence of lamellar order in intermix region, similar to previous observation (Fig. 2c, Supplementary Figs. S25, S26)¹¹. Therefore, a key progress within transition is the formation of short-range aggregates of amorphous acceptor (quasi-aggregates), which creates a special quasi-aggregated intermix region, containing the interfacial characteristics. The weaker miscibility of SMA with PD leads to weak percolation, constraining the growth of quasi-aggregates but promoting the growth of pure region^24,41. In contrast, the stronger interfacial percolation in the PA system promotes larger quasi-aggregated intermix region and more robust polymer network for charge transport (Fig. 2c)⁴². The quasi-aggregates in intermix region can be uniformly observed across representative seven OPV systems (Supplementary Figs. S25–S27). Here, we clarified the entangled amorphous regions and quasi-aggregated regions as two distinct interfaces: (1) The Entangled-interface (E-interface), a fully amorphous (isotropic) region where donor and acceptor are completely entangled (cautiously modeled by the feature of 10-1 D-A ratio); and (2) The Penetrated-interface (P-interface), formed by the crystallized quasi-aggregates (short-range aggregates of amorphous acceptor) and those penetrated PD chains (cautiously modeled by the feature of 10-15 D-A ratio). These distinctions elucidate the structural evolution of phase separation across different D-A compositions and systems. In particular, PA system exhibits a gradual transition from the amorphous E-interface to the dominated quasi-ordered P-interface, and ultimately to the ordered crystalline pure phase. We propose to utilize ΔrDoC to qualitatively determine the population of P-interface. As a results, PA system presents a more substantial P-interface population (larger ΔrDoC). In sharp contrast, the rapid transition of SMA system indicates the limited development of both E- and P-interface (small ΔrDoC), with the dominated former. We suspect that the formation of isolated acceptor units (or tiny clusters) in isotropic intermix region may lead to the deep traps at E-interface, while the formation of short-range quasi-aggregates at P-interface may supress those deep traps, which promote the energetic order^43,44.

As established, excitons dissociate into charge-transfer (CT) states and then separate into free carriers at interfaces, underscoring the crucial need to differentiate the morphology of E- and P-interfaces in both PA and SMA systems^45,46. We quantitatively analyze the non-crystalline ring associated with the interface (in the range of 1.2–1.5 A⁻¹). Given the amorphous nature of the E-interface, we hypothesize that this packing feature contributes to the quasi-crystalline P-interface. Thus, we analyze the morphology of the P-interface by deconvoluting IP scattering intensities, especially focusing on edge-on orientation (detail see Supplementary Fig. S28). The edge-on ratio (intensity ratio of quasi-aggregates and lamellar stacking) in the PA and SMA system (Fig. 2d) follows the transitional trends the growth of the P-interface. This observation indicates that the P-interface, while still limited, exhibits a marginally greater degree of edge-on features than the amorphous E-interface. The comparatively more apparent edge-on characteristics are observed in the PA system, cautiously supporting by higher electron mobility and its faster evolution in OFET devices (Fig. 2d, Supplementary Fig. S24). We propose the crossing structure for P-interface exhibiting unaligned and tilted packing between donor and acceptor, and the parallel structure for E-interface with better molecular alignment (Supplementary Fig. S29). Notably, we emphasise that although the P-interface exhibits more pronounced edge-on features than pure phase, it is still significantly dominated by face-on orientation.

We, subsequently, performed GISAXS measurements to determine the size of the intermix region (ξ_interface), containing both E-interface and P-interface, and the crystalline pure domain (2R_g-Acceptor) (see Supplementary Note S1)³⁹. The sizes of the intermix domains decrease and the pure domain increase with increase of D-A ratio. PA system shows larger interface domain, while SMA system shows larger pure domain (Supplementary Fig. S30), aligning with GIWAXS results, indicating stronger preference of forming interface in PA systems. Overall, these morphological results confirm the existence of the P-interface in both the PA and SMA systems. The PA system exhibits the more pronounced P-interface with more obvious edge-on features and larger intermix domains than the SMA system. In contrast, the P-interface in the SMA system evolves more rapidly despite its lower population. These distinct morphologies significantly influence the optoelectronic properties of CT states, which can be estimated by theoretical simulation and calculation⁴⁷.

Theoretical simulation and calculation

According to the results of molecular dynamics simulation, we observed that the donor and acceptor molecules exhibit different stacking arrangements depending on the large-scale molecular organization. For analytical clarity, we adopt prior definitions of two primary configurations: parallel and cross structure, which are consistent with our prior analysis of the two interfacial morphologies (Fig. 3a). When donor–acceptor dihedral angles are small (0°–45° for PA system, and 0°–60° for SMA system), molecules adopt parallel stacking structure (E-interface), while larger angles (75–130° for PA system and 80°–150° for SMA system) yield cross-stacking structure (P-interface).

Fig. 3. Simulated illustration for P- and E-interface of two systems.

a The MD illustration for the shape of E- and P-interface with parallel and crossing structures. b Calculated Huang–Rhys factor for two systems with varying dihedral angle. c Schematic diagram of the potential energy curves for the S₀ states, CT states for E- and P-interfaces, S₁ diabatic states of PA and SMA systems with their reorganization energy.

Dihedral angle distributions under varying D-A ratios show a systematic shift from parallel to cross-stacking with increasing D-A ratio, correlating with P-interface formation (Supplementary Fig. S31). This transition reflects disrupted alignment between donor chains and acceptors upon P-interface growth. The trend is particularly pronounced in PA systems, where P-interfaces dominate more distinctly than in SMA systems (Supplementary Fig. S31b). Both interface types and molecular systems establish unique equilibrium geometries and excited-state potential energy surfaces, including interfacial charge-transfer (CT) states and molecular S₁ states³. The electronic transitions between S₁ state and CT states induce the coupling between intra-molecular π-electron redistribution (intramolecular conformational changes) and the molecule lattice (phonon) vibration, and the intermolecular reorganization of nuclear position and charge distribution during lattice relaxations. Critically, we calculate the inner reorganization energy ${\lambda }_i$, describing the energy for the molecular lattice distortion, to quantify the cost energy for intra-molecular geometric changes when charge transfer⁴⁸. For the D18-PYIT system, both the E-interface and the P-interface exhibit lower ${\lambda }_i$, suggesting minimized potential energy changes between ground and excited states in polymer acceptors (Supplementary Fig. S32). The PA systems show small ${\lambda }_i$ may attributed to its greater delocalized π-electron density distribution from the longer effective conjugation lengths of PA relative to SMA (Supplementary Figs. S33, S34)²⁴. We then calculate the outer reorganization energy ${\lambda }_o$ for both systems by fitting the reduced electroluminance (rEL) spectra, referred by PL spectra under short-circuit condition, according to modified classical Marcus theory approach (Supplementary Figs. S35–S38, Supplementary Tables S4, S5 and Supplementary Note S2). CT state energy (E_CT) can also be extracted for varied D-A ratio devices of two systems. The observed smaller ${\lambda }_o$ indicates the more ordered interface in PA systems⁶. E_CT continuously decreases with increasing acceptor ratio, consistent with the growing presence of the P-interface. The reduction in reorganization energy ($\lambda$, including ${\lambda }_i$ and ${\lambda }_o$) is beneficial for minimizing the driving force for exciton dissociation by mitigating molecular distortion energy penalties⁴⁹.

Within the Marcus framework, the non-radiative decay of excited state is strongly influenced by the molecular reorganization, describing the electron-phonon coupling (EPC) induced by molecular vibration and the ubiquitous disorder (Supplementary Note S3). We quantify the EPC by Huang–Rhys factor, S, assess how distinct molecular arrangements affect reorganization energy and non-radiative recombination:

$$S=\frac{{\lambda }_i}{\hslash \Omega }$$ 2

Where the ${\lambda }_i$ is the inner reorganization energy and ℏΩ represents the energy difference between vibrational modes². During electronic transitions, lattice equilibrium shifts break phonon mode orthogonality, enabling more phonons to participate in the transition process. This process, often accompanied by phonon excitation, enhances non-radiative recombination probability. A higher S value indicates stronger EPC and greater non-radiative transition probability, typically originating from intensified intramolecular vibrational coupling.

According to our classification of molecular stacking, parallel configurations dominate E-interfaces, while cross-stacking prevails at the P-interface. Simulations reveal higher EPC in SMA versus PA systems for both interface types and, in addition, the EPC of E-interface is also higher than P-interface in both systems. The result of smaller EPC of P-interface aligns with the suppression of EPC by increasing the D-A spacing⁵⁰, where the spacing of P-interface (larger dihedral angle) is larger than E-interface (smaller dihedral angle; Fig. 3b). There are two key factors drive the enhanced EPC in SMA systems, including lattice displacement and different-frequency vibrational modes. SMAs undergo greater lattice displacement during electronic transitions, resulting in a higher ${\lambda }_i$ (Supplementary Fig. S32). These vibrational interactions amplify EPC^51,52. The suppression of EPC occurs in packed aggregates due to restricted backbone conformational changes during charge transfer, with further reduction from restricted vibration coupling and increased electron delocalization, which indicates the reason of weaker EPC of P-interface^48,49. The PA systems likely show weaker EPC due to the more delocalized nature of electrons in PA. Experimentally, we conform the S-value by determining the shifts of Barycentre (ΔE_BC) in absorption and emission spectra (ΔE_BC = 2S) (Supplementary Fig. S39, Supplementary Note S4). The calculated S value and trends show good agreement with simulated results. We further calculate the S value for seven systems (Supplementary Fig. S40), including four SMA systems and three PA systems, where SMA systems always exhibit higher value than PA systems, thereby ensuring the validation across the state-of-the-art OPV systems.

To further investigate the impact of P-interfaces on CT and singlet excited (S₁) state interactions, we performed time-dependent density functional theory (TD-DFT) calculations for both interface types. As shown in Supplementary Figs. S41, S42, the electronic coupling between the S₁ and CT states ($t_{{\mathrm{S}}_1-\mathrm{CT}}$) was found to be consistently higher at the E-interface than at the P-interface. Specifically, For D18-PYIT, $t_{{\mathrm{S}}_1-\mathrm{CT},\mathrm{PA}-\mathrm{P}}$ of P-interface is averaged 2.5 meV (max 14 meV) versus $t_{{\mathrm{S}}_1-\mathrm{CT},\mathrm{PA}-\mathrm{E}}$ of 2.7 meV (max 34 meV) at E-interface. Similarly, D18-eC9 showed $t_{{\mathrm{S}}_1-\mathrm{CT},\mathrm{SMA}-\mathrm{P}}$ of 3.3 meV (max 18 meV) at P-interfaces versus $t_{{\mathrm{S}}_1-\mathrm{CT},\mathrm{SMA}-\mathrm{E}}$ of 4.2 meV (max 40 meV) at E-interfaces. The trends of those $t_{{\mathrm{S}}_1-\mathrm{CT}}$ might be corroborated experimentally by charge transfer rates based on Marcus Theory. Critically, both interfaces exhibit comparable coupling magnitudes, potentially enabling the dual pathways for exciton dissociating into free charges. These comparable $t_{{\mathrm{S}}_1-\mathrm{CT}}$ promote the hybridization between CT and S₁ to reduce non-radiative recombination energy losses. However, in addition to hybridization, the significantly distinct λ in two systems (Fig. 3c), forcing variable EPC and electron transfer rate, may also substantially influence charge carrier dynamics and qΔV_nr, according to Marcus Theory.

Charge generation dynamics with the effects of P-interface

We perform Pump-Probe (PP) and Pump-Push-Probe (PPP) transient absorption spectroscopy (TAS) and to investigate charge generation dynamics at the interface and trace the population dynamics of bound CTEs (Supplementary Figs. S43–S56)^53,54. The different morphology between P-interface and E-interface affects the exciton dissociation and also the CTE separation. We analyze the hole transfer kinetics of donor ground-state bleach (GSB) signal by tri-exponential function to analyze the charge generation process (Fig. 4a, Supplementary Figs. S44, 45, Supplementary Table S6, and Supplementary Note S5).

Fig. 4. Charge carrier kinetics and illustrations.

a Hole transfer kinetics from PP-TAS of PA and SMA systems fitting by tri-exponential function, where ${\tau }_n$ present the time constant of two interfaces and diffusion-limited transfer, and A_n are the weight coefficient for each term. b The ratio of A₁ and A₂ for two systems. c Bound CTEs separation kinetics from PPP-TAS at representative EA response wavelength for PA and SMA system. d Illustration of the dual charge generation pathways originating from the E-and P-interfaces, highlighting their distinct population ratios and charge transfer time. Gray shaded shape on (a) represents the laser pulse width of transient absorption experiments showing the high time resolution here.

The initially formatted E-interface promotes early charge generation and dominates the hole transfer process, especially at low D-A ratios (10-1) (Fig. 4a, light red and light blue). The following P-interface formation and expansion significantly impacts the subsequent charge generation as the ratio increases (10-15) (Fig. 4a, dark red and dark blue). The evolution of the E-interface (${\mathrm{A}}_1$) and P-interface (${\mathrm{A}}_2$) charge transfer weight coefficient, quantifying the ratio of two interfaces, aligns with the growth trends of the P-interface in two systems (Fig. 4b). The population of P-interface in PA system is larger than that in E-interface and grows continuously, while in SMA system, the E-interface dominance the interface population. We argue that charge generation occurs via both interfaces, E-interface (modeled at a 10-1 D-A ratio), and P-interface (modeled at a 10-15 D-A ratio) (Supplementary Fig. S46). The faster ${\tau }_1$ and ${\tau }_2$ in the PA system likely originate from more favorable values of $\mathrm{\Delta }{E}_{{\mathrm{S}}_1\text{-}\mathrm{CT}}$ and $\lambda$. The smaller difference between ΔE_S1-CT and $\lambda$ leads to faster charge transfer rate according to classical Marcus theory, facilitating faster exciton dissociation and CTEs generation⁵⁵. We further estimated the electronic couplings between the S₁ and CT states for the two interfaces, $t_{{\mathrm{S}}_1-\mathrm{CT},\mathrm{E}}$ and $t_{{\mathrm{S}}_1-\mathrm{CT},\mathrm{P}}$ (Supplementary Table S7). The estimated values agree with the trends from theoretical calculation and offer valuable insights into the degree of hybridization between the two states². The diffusion-limited transfer time constant, ${\tau }_3$, and its weight coefficient, ${\mathrm{A}}_3$, increased significantly due to the growth of pure domain sizes, leading to longer diffusion times⁵⁶. The shorter ${\tau }_3$ in PA system might due to the more gentle energy landscape between pure phase and interface, facilitating exciton migration⁶.

After studying the charge transfer process, we then examined CTE separation using PPP-TAS measurements, which directly probe this process (Supplementary Fig. S48 and Supplementary Note S5)⁵⁷. Fig. 4c presents the kinetics of the push response aligning with the electro-absorption (EA) response at 610 nm (Supplementary Fig. S49), which confirms the presence of push-induced CTE states⁵³. The lower normalized PPP response intensity in the PA system, which quantifies the bound CTE population, indicates a minimally retained CTE population and, therefore, efficient CTE separation. Besides, the faster decay of the PPP response kinetics in the PA system suggests the more weakly bound CTEs, facilitating their separation into free charges. This may be attributed to the more delocalized orbitals overlapping of PA system^58,59. This observation is consistent with the shorter charge generation times (τ₁ and τ₂), highlighting the faster CTE separation dynamics in the PA system.

Throughout the charge generation process, the PA system manifests a gentler energy landscape that facilitates faster exciton diffusion and an energetically ordered interface promoting exciton dissociation⁶. Our PPP-TAS results also demonstrate faster charge-transfer exciton (CTE) separation rates in PA system relative to SMA system. Besides, the growing population of P-interface assists the exciton dissociation in both systems. We proposed a dual pathway for charge generation originating from E- and P-interface, wherein latter may facilitate fast and efficient CTEs separation (Fig. 4d)⁶⁰. The different morphologies and charge generation characteristics of the P-interface highlight the distinct electronic properties in the two systems.

Non-radiative energy loss from E- and P-interfaces

To assess the impact of P-interface on device performance, we fabricated additive-free devices to isolate performance variations arising solely from D-A interactions. Besides, the elected acceptors in SMA systems and PA systems possess identical HUMO levels and share the same donor, eliminating the driven force induced energy loss differences. Thus, observed energy loss disparities primarily originate from multiple CT states⁶¹. As discussed in charge generation section, E_CT exhibits distinct trends in two systems, where PA system shows gradual reduction of E_CT with increasing acceptor content, while SMA system exhibits sharp drops of E_CT until the D-A ratio threshold. Analysis of the energy loss, E_loss, (Supplementary Fig. S57, Supplementary Table S8 and Supplementary Note S6) reveals significantly lower losses in PA system than SMA system, particularly the qΔV_nr of ~ 63 meV smaller, as confirmed by EQE_EL (Fig. 5a and Supplementary Fig. S58)²⁶. In the PA system, qΔV_nr decrease monotonically with D-A ratio increase, whereas SMA system shows a decline until 10-5 ratio followed by stabilization-consistent with prior observations. The employment of different transport layers confirms these decreasing trends, indicating that the energy loss arises not from the interlayers but from the interfaces of the active layer (Supplementary Figs. S59, S60). Critically, the key difference in energy loss between PA and SMA systems stems from qΔV_nr, governed by k_r, and k_nr.

Fig. 5. Energy loss and non-radiative energy loss analysis by simulation and experiments.

a E_loss and qΔV_nr of PA and SMA system with increased D-A ratio. b The simulated $k_{\mathrm{r}}$ and $k_{\mathrm{nr}}$, and c corresponding simulated qΔV_nr are plotted for the cases of low (red), moderate (yellow), and high S (blue) as a function of $\mathrm{\Delta }{E}_{{\mathrm{S}}_1-\mathrm{CT}}$ by considering the empirical parameters. d The simulated $k_{\mathrm{r}}$ and $k_{\mathrm{nr}}$ as a function of S by considering the experimental parameters. e The expected qΔV_nr for P- and E-interface of two systems as a function of S, and the comparison to experimental qΔV_nr from devices. f J–V curve of the optimal devices based on D18-PYIT-eC9 with ratio of 1-0.2-1 according to the pseudo-bulk heterojunction (layer-by-layer) ternary strategy.

To investigate the difference between PA and SMA, and how the presence of P-interface impacts qΔV_nr, we estimate k_r and k_nr, via the three-state model. Unlike the conventional two-state model, which only accounts for non-radiative energy loss between CT and S₀ states, the three-state model additionally incorporates hybridization between S₁ and CT states, enabling comprehensive recombination loss assessment^6,9. The k_r is derived from Einstein’s spontaneous emission equation, while the k_nr is determined using the modified Marcus–Levich–Jortner (MLJ) formula (Supplementary Note S7)²:

$$\mathrm{FCWD}\left(E\right)\propto \sum _{w=0}^{\infty }\sum _{l=0}^{\infty }\frac{e^{-S}{S}^{w-l}l!}{w!}{e}^{-\left\{\frac{{\left[\mathrm{E}-{\mathrm{E}}_{\mathrm{CT}-{\mathrm{S}}_0}+{\mathrm{\lambda }}_{\mathrm{o}}+\left(w-l\right)*\hslash \Omega \right]}^2}{4{\lambda }_o{k}_{\mathrm{B}}T}\right\}}$$ 3

$$k_{\mathrm{r},\mathrm{CT}-{\mathrm{S}}_0}=\frac{{\left(E_{\mathrm{CT}-{\mathrm{S}}_0}\right)}^3}{3\pi {ϵ}_0{\hslash }^4{c}^3}{\left(∣\overrightarrow{{\mu }_{\mathrm{CT}-{\mathrm{S}}_0}}∣+\frac{∣\overrightarrow{{\mu }_{{\mathrm{S}}_1-{\mathrm{S}}_0}}∣*t_{{\mathrm{S}}_1-\mathrm{CT}}}{\mathrm{\Delta }{E}_{{\mathrm{S}}_1-\mathrm{CT}}}\right)}^2\mathrm{FCWD}\left(E_{\mathrm{CT}-{\mathrm{S}}_0}\right)$$ 4

$$k_{\mathrm{nr},\mathrm{CT}-{\mathrm{S}}_0}=\frac{2\pi }{\hslash }{\left(t_{\mathrm{CT}-{\mathrm{S}}_0}+\frac{t_{{\mathrm{S}}_1-{\mathrm{S}}_0}{t}_{{\mathrm{S}}_1-\mathrm{CT}}}{△E_{{\mathrm{S}}_1-\mathrm{CT}}}\right)}^2\mathrm{FCWD}\left(0\right)$$ 5

Where Franck–Condon weighted density of states (FCWD) accounts for transitions between all vibrational states of CT states and ground states; $E_{\mathrm{CT}-{\mathrm{S}}_0}$ is the energy difference between the CT state and the S₀ state during electron transfer; w and l denote the quantum number of the vibrational modes of the CT state and the S₀ state, respectively; ${\lambda }_o$ is associated with the outer reorganization energy; $\overrightarrow{{\mu }_{{\mathrm{S}}_1-{\mathrm{S}}_0}}$ and $\overrightarrow{{\mu }_{\mathrm{CT}-{\mathrm{S}}_0}}$ are the transition dipole moment between the S₁ state to S₀ states, and the CT state to ground states, respectively; ${ϵ}_0$ is the vacuum permittivity; ℏ is the reduced Planck constant; and c is the speed of light in vacuum; $t_{{\mathrm{S}}_1-{\mathrm{S}}_0}$, $t_{\mathrm{CT}-{\mathrm{S}}_0}$ and $t_{{\mathrm{S}}_1-\mathrm{CT}}$ are electronic coupling between the S₁ state to ground states, the CT state to S₀ states, and the S₁ state to CT state, respectively; S is the Huang–Rhys factor that used to quantify the extent of EPC^1,2,4.

Both $k_{\mathrm{r},\mathrm{CT}-{\mathrm{S}}_0}$ and $k_{\mathrm{nr},\mathrm{CT}-{\mathrm{S}}_0}$ depend on the EPC, but more strike on $k_{\mathrm{nr},\mathrm{CT}-{\mathrm{S}}_0}$. We calculate $k_{\mathrm{r},\mathrm{CT}-{\mathrm{S}}_0}$, $k_{\mathrm{nr},\mathrm{CT}-{\mathrm{S}}_0}$ and ΔqV_nr as a function of $\mathrm{\Delta }{E}_{{{\mathrm{S}}_1}_{\text{-}}CT}$ for the three cases of small S (0.8), moderate S (1), and huge S (1.2) using the parameter in Supplementary Table S9. As shown in Fig. 5b, it is clear that $k_{\mathrm{r},\mathrm{CT}-{\mathrm{S}}_0}$ shows relatively marginal variation with the change of S, while $k_{\mathrm{nr},\mathrm{CT}-{\mathrm{S}}_0}$ exhibits a significant dependence on S, demonstrating the dominant role of latter in reducing qΔV_nr. Consequently, the simulated qΔV_nr decreases about 90 meV as S increased from 1.2 to 0.8, where $k_{\mathrm{nr},\mathrm{CT}-{\mathrm{S}}_0}$ decrease significantly (by orders of magnitude), while $k_{\mathrm{r},\mathrm{CT}-{\mathrm{S}}_0}$ decreases slightly (less than one order) (Fig. 5c). Furthermore, this ~ 90 meV EPC-induced reduction of energy loss is largely independent of $△E_{{\mathrm{S}}_1-\mathrm{CT}}$, especially across 20–100 meV (matching the experimental $△E_{{\mathrm{S}}_1-\mathrm{CT}}$). Overall, reducing S in low-offset systems can therefore mitigate the non-radiative energy losses¹⁶.

Beyond EPC, prior studies have identified that parameters such as, $△E_{{\mathrm{S}}_1-\mathrm{CT}}$, $E_{\mathrm{CT}}$, $t_{{\mathrm{S}}_1-\mathrm{CT}}$, and ${\lambda }_o$, also contribute to the reduction of qΔV_nr^45,49. It is therefore crucial to distinguish the individual contribution of each parameter to qΔV_nr. Regardless of the EPC, the reduction in qΔV_nr achieved by lowering ${\lambda }_o$ is independent of the change of $△E_{{\mathrm{S}}_1-\mathrm{CT}}$, whereas V_OC increases by varying the $t_{{\mathrm{S}}_1-\mathrm{CT}}$, $t_{{\mathrm{S}}_1-{\mathrm{S}}_0}$, and $t_{\mathrm{CT}-{\mathrm{S}}_0}$ are as the functions of $△E_{{\mathrm{S}}_1-\mathrm{CT}}$ (Supplementary Fig. S61 and Supplementary Tables S10, 11). Specifically, the V_OC increase by reducing ${\lambda }_o$, from 200 meV to 100 meV, remains consistent, as ~85 meV. Conversely, the positive impact of modulating electronic couplings (or transition dipole moment) on V_OC is gradually diminishes as $△E_{{\mathrm{S}}_1-\mathrm{CT}}$ decreases. However, this diminishing effect could be counteracted by reducing EPC which is isolated from $△E_{{\mathrm{S}}_1-\mathrm{CT}}$ (Supplementary Fig. S62 and Supplementary Tables S12, 13). Such a reduction of EPC might further enhance the molecular PLQY and EQE_EL⁶², and is also independent of $△E_{{\mathrm{S}}_1-\mathrm{CT}}$, presenting a potential strategy for supressing non-radiative energy loss of OSCs.

The non-radiative energy losses in the two systems were analyzed following the insights from numerical simulations. Calculation of the recombination rates, $k_{\mathrm{r},\mathrm{CT}-{\mathrm{S}}_0}$, $k_{\mathrm{nr},\mathrm{CT}-{\mathrm{S}}_0}$, and qΔV_nr as functions of EPC based on the experimental parameters (Supplementary Table S7) show a consistent ~28 meV lower qΔV_nr for P-interface dominant devices (D-A = 10–15) compared to E-interface dominant ones (D-A = 10-1) across both PA and SMA systems (Fig. 5d, e, Supplementary Fig. S63). Attribution of this reduction points to the smaller reorganization energy and stronger hybridization effects of P-interface. Decomposition of the contributions confirm ${\lambda }_o$ as the dominant factor, consistent with simulations (Supplementary Fig. S64). It is notable that the reducing of $△E_{{\mathrm{S}}_1-\mathrm{CT}}$ might enhance both the electronic part of $k_{\mathrm{r}}$ and $k_{\mathrm{nr}}$, making careful balancing of them essential. Similarly, replacing SMA with PA reduces qΔV_nr by ~25 meV also due to the smaller $△E_{{\mathrm{S}}_1-\mathrm{CT}}$ and larger $t_{{\mathrm{S}}_1-\mathrm{CT}}$ leading to enhanced S₁-CT state hybridization, and smaller ${\lambda }_o$.

These results agree with the established protocols for reducing qΔV_nr by decreasing ${\lambda }_o$ (reducing FCWD), and increasing the degree of hybridization, achieving by the formation of the P-interface and the utilization of PA. However, the achieved reductions (~25 meV and ~28 meV) are unable to fully account for the observed qΔV_nr in devices with varying D-A ratios (~55 meV for PA system and ~38 meV for SMA system between 10-1 and 10-15 devices) and replacing SMA to PA system (~63 meV between 10-15 devices). We then consider examining the role of EPC in mitigating qΔV_nr. Consistent with predictions, $k_{\mathrm{nr},\mathrm{CT}-{\mathrm{S}}_0}$ decreases dramatically but $k_{\mathrm{r},\mathrm{CT}-{\mathrm{S}}_0}$ increases slightly with the reduction of S. The former contributes the most for reducing qΔV_nr.

TD-DFT simulations show that in PA systems, S evolves from 1.04 (E-interface) to 0.83 (P-interface), resulting in a reduction of qΔV_nr by ~58 meV when considering the population weight of E- and P-interface at varying D-A ratio from 10-1 to 10-15 (${\mathrm{A}}_{\mathrm{E}}=\frac{{\mathrm{A}}_1}{{\mathrm{A}}_1+{\mathrm{A}}_2}\mathrm{and}{\mathrm{A}}_{\mathrm{P}}=\frac{A_2}{A_1+A_2}$, Supplementary Fig. S65). This closely matches the experimental reduction (~55 meV). Similarly, in the SMA system, theoretical predictions of EPC reduction (S = 1.11 for E-interface and S = 0.95 for P-interface) suggests a ~37 meV decrease in qΔV_nr, aligning well with experimental results ($~$38 meV). The smaller reduction of qΔV_nr in the SMA system is attributed to the less population of P-interface. Besides, the reduction of qΔV_nr predicted by experimental EPC also agree with the qΔV_nr results of devices. Ignoring parameter except EPC, we exhibit a ~35 meV and ~20 meV reduction of qΔV_nr for PA and SMA systems, respectively, which all contribute more than half of the total loss. These findings further support the strong correlation between the increase of P-interface and the reduction of qΔV_nr.

We correlate the experimental energy loss values with simulated energy loss, inferring that both the PA and SMA systems may exhibit diminished EPC, coinciding with the gradual growth of the P-interface. Notably, the more pronounced P-interface in PA system supresses more in energy loss (~55 meV of PA system and ~38 meV of SMA system). This underscores the importance of suppressing EPC by introducing more P-interface, which inherently possesses lower EPC. Additionally, the recent PA systems exhibit a closer proximity between the S₁ and CT states, resulting in stronger hybridization and further reduction of non-radiative losses². In addition to the representatives of PA system, D18-PY-IT, and SMA system, D18-eC9, other two PA systems also exhibit larger P-interface population and smaller EPC than those of three SMA systems (Fig. 6a, b). We summarize the many of the reported energy losses in literatures achieved by two systems (Fig. 6c). It is obvious that PA-based systems show potentially lower ΔV_nr than SMA systems. We propose achieving low ΔV_nr in binary system by modulating the EPC of the interface, particularly by leveraging the P-interface with inherently lower EPC, which benefits from strong percolation between polymer acceptor and polymer donor.

Fig. 6. Mechanism diagram and universality of more pronounced P-interface with weaker EPC reducing non-radiative voltage loss in OPVs.

a The calculated S factor value of SMA, PA and ternary systems. b The calculated ΔrDoC of SMA, PA and Ternary systems. c The summarization of many reported energy losses in literatures about two systems. d The photogenerated excitons diffuse from pure phase to interface, following the exciton dissociation to CTEs at E-interface (intermix region, deep red bracket covered region) and P-interface (transition region filled by quasi-aggregates and penetrated polymer chain or molecules, light red bracket covered region), and finally CTEs separation to free charges.

In fact, we argue that the state-of-art strategies implicitly imply the existence of the P-interface which ultimately optimizes the V_OC of the device. These strategies include the introducing the third component for regulating interfacial morphology⁶³, employing the (diluted) pseudo-bulk heterojunction (layer-by-layer) strategy to modify the intermixing regions^43,64,65, and using the additive to both donor and acceptor domains^15,66,67. All these strategies can be well described by design rules of increasing the emission of CT states or S₁ states⁶⁸. Based on these insights, we process the ternary system, D18-PYIT-eC9, using the pseudo-bulk heterojunction strategy to form a p-i-n structure that the incorporation of PD modulates the population of P-interface in ternary device with the comparison to its PD-SMA binary counterpart. The most optimized device achieves 20 meV increase in V_OC and over 18% efficiency, with no sacrifice in J_SC and FF, and without the need for additives or other special treatments (Fig. 5f). We cautiously attribute this increase to the increased population of P-interface and then the suppression of EPC (Fig. 6a, b), facilitating more emissive CT states and minimized non-radiative energy losses.

Discussion

We have conducted a systematic investigation of both polymer-NFA (SMA system) and all-polymer (PA system) blends, with a focus on the evolution of interfacial morphology, the enhancement of charge generation efficiency, and the suppression of non-radiative energy loss. These improvements are attributed to the formation of the P-interface, which exhibits both morphological and energetic order, along with inherently weaker electron–phonon coupling (EPC) (Fig. 6a). We attribute the formation of P-interface to the growth of quasi-aggregates existed in intermix region, quantifying by the value of ΔrDoC (Fig. 6b). Through simulations, GIWAXS/GISAXS, and transient absorption spectroscopy (TAS) measurements of on blends with varying acceptor concentrations, we demonstrate that continuous incorporation of acceptor into donor matrix promotes the formation of E- and P-interface. Notably, the P-interface becomes more pronounced in the PA system due to stronger interfacial percolation, while the E-interface is more dominant in the SMA system. The calculated charge transfer rate coefficient ratio of P- and E-interface (${\mathrm{A}}_2/{\mathrm{A}}_1$) provides the straightforward relation of the population of two interface. In the PA system, the well-developed P-interface not only introduces a second pathway for efficient charge generation but also significantly suppresses EPC, leading to reduced non-radiative recombination rates, $k_{\mathrm{nr}}$, and substantial decrease in non-radiative energy loss, qΔV_nr. In contrast, the SMA system, with a lower P-interface population, limits the channel for charge generation and retains more CT excitons (CTEs) at the interface. Additionally, the prevalence of the E-interface with stronger EPC in SMA system contributes to higher $k_{\mathrm{nr}}$ and greater qΔV_nr.

In conclusion, the observed suppression in voltage loss between all-polymer solar cells and Y-series SMA-based solar cells, ~60 meV, is closely associated with the more pronounced P-interface, which exhibits suppressed electron–phonon coupling (EPC) within the polymer–polymer matrix of all-polymer systems. Specifically, In PA system, the more pronounced P-interface, formed by the quasi-aggregates, constructs a gentle energy landscape that facilitates the efficient charge generation, while the steeper energy landscape in the SMA system hinders the charge generation dynamics. We propose that the relative population of E- and P-interface can be estimated by calculating ΔrDoC and ${\mathrm{A}}_2/{\mathrm{A}}_1$. Subsequently, in the PA system, more pronounced P-interfaces contribute to a notable qΔV_nr reduction of ~55 meV (Fig. 6), with weaker EPC at the P-interface responsible for over half of this value (~35 meV). In comparison, the SMA system, with less proportion of P-interfaces, exhibits a smaller suppression of qΔV_nr (~38 meV), yet the EPC-related contribution remains substantial (~20 meV), accounting for approximately half of the total suppression. Guided by these insights, we fabricate the D18-PYIT-eC9 ternary devices by pseudo-bulk heterojunction strategy, achieving increased V_OC of ~20 meV. This emphasizes the importance of properly optimizing the growth of P-interface as well as the energetic landscape from bulk to interface for the further development of OSC performance.

Methods

Device fabrication and testing

Solar cells were fabricated in a conventional configuration of ITO/PEDOT:PSS/active layers/PFN-Br/Ag. The ITO substrates were first scrubbed by detergent and then sonicated with deionized water, acetone and isopropanol subsequently, and dried overnight in an oven. The glass substrates were treated by UV-Ozone for 30 min before use. PEDOT:PSS (Al4083 from Hareus) was spin-cast onto the ITO substrates at 7500 rpm for 30 s, and then dried in ambient atmosphere. The blend solution was spin-cast at 2000–2500 rpm for 30 s onto PEDOT:PSS film followed by a temperature annealing of 100 °C for 1 min. A thin PFN-Br-MA layer (0.5 mg/mL in methanol and 0.25% wt% melamine, 3000 rpm) was coated on the active layer, followed by the deposition of Ag (evaporated under 3 × 10⁻⁴Pa through a shadow mask). The current density-voltage (J–V) curves of devices were measured using a Keysight B2901A Source Meter in glove box under AM 1.5 G (100 mW cE-²) using an Enlitech solar simulator. The device contact area was 0.042 cm², device illuminated area during testing was 0.041 cm², which was determined by a mask. A step voltage of 0.1–0.5 V with about 5 ms delay time are used. Device measurements and thermal ageing were conducted inside the glovebox under an inert atmosphere.

GIWAXS and GISAXS

2D GISAXS/GIWAXS measurement was performed on an XEUSS 3.0 UHR SAXS/WAXS system (XENOCS, France). A Eiger2 R 1 M 2-dimensional detector with 0.075 mm × 0.075 mm active pixels were utilized in integration mode. The sample-to-detector distance is settled at 100/2000 mm for GIWAXS/GISAXS measurement. The precise sample-to-detector distance was determined with a silver behenate standard. The Cu incident X-ray (8 KeV) with a 0.9 mm × 0.9 mm/0.5 mm × 0.5 mm spot provided large enough q space. 1D GIWAXS patterns was corrected to represent real q_r and q_z axis with the consideration of missing wedge. The critical incident angle was determined by the maximized scattering intensity from sample scattering with negligible contribution from underneath layer scattering. The incident angle scattering was collected at 0.2°, which renders the incident X-ray as an evanescent wave along the top surface of thin films. The samples for GIWAXS/GISAXS test were prepared by casting solution onto silicon wafer substrates (ca. 15 mm × 15 mm), and the active layers were prepared using exactly the same concentration and same procedures as those for J–V measurements.

PP- and PPP-TAS

PP-Transient absorption measurement was conducted on a commercial pump-probe femtosecond transient absorption (TA) spectrometer Helios (Ultrafast System, USA). Ultrafast laser pulses (800 nm, <35 fs pulse duration, 7 W) was generated by 1 kHz Ti:Sapphire regenerative amplifier (Astrella, Coherent, USA). 40% of the fundamental pulses (7 W) was used to pump the commercial collinear optical parametric amplifier (TOPAS Prime, Light-Conversion LLC, Lithuania) for generating tunable wavelength pump pulse to 400 nm or 800 nm. The pump beam is chopped at 500 Hz. Another 45% of the fundamental pulses (7 W) is used to pump another the commercial collinear optical parametric amplifier (TOPAS Prime, Light-Conversion LLC, Lithuania) for generating 2000 nm wavelength push pulse. The push beam is chopped at 250 Hz. The push beam was been delay related to pump beam by a commertial delay stage. Rest 15% of the fundamental pulses was routed onto a mechanical delay stage (within 7 ns) and passed through a sapphire crystal to generate supercontinuum probe light (450–750 nm) and a YAG crystal to generate NIR supercontinuum probe light (800–1200 nm). The pump light and probe light were focused on a same spot (~2 mm diameter) of the thin films placed on a quartz. Data analysis is performed by Surface Xplorer software. The incident power is measured with a calibrated laser power meter (Newport).

Reporting summary

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

Author contributions

Y.L. (Y. Luo), Y.H. and Y.L. (Y. Li) contributed equally. J.W. and Y.L. (Y. Luo) conceived the project and J.W. supervised this work. OSC samples were prepared by R.M., and L.W., under the supervision of G.L., and T.J.; Transient Absorption experiments (PP-TAS) were carried out by Y.L. (Y. Luo) and Y.L. (Y. Li); Pump-push-probe TAS (PPP-TAS) and in-situ EPC were carried out by Y.L. (Y. Li) and Y.L. (Y. Luo); GIWAXS/GISAXS measurements were carried out and analyzed by Y.L. (Y. Luo); simulated calculation was carried out by Y.H. supervised by J.W. OFET devices fabrication and measurements were carried out by F.D. and M.W. Streak Camera-based TRPL measurements are carried out by Y.C. T.A.D.P., H.Y. and Y.M.L. helped to analyze the data. J.W., G.L., T.J. and H.Y. provided experiment conditions. Y.L. (Y. Luo), Y.H., Y.L. (Y. Li) and J.W. wrote the manuscript. All authors commented on the manuscript.

Peer review

Peer review information

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

Data availability

All data supporting the findings of this study are available within the main text and Supplementary Information file. Additional data are available from the corresponding author on request. 10.6084/m9.figshare.29390669.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68731-7.